Author's response to reviews Title:DNA Methylation Subgroups and the CpG Island Methylator Phenotype in Gastric Cancer: A Comprehensive Profiling Approach Authors: Marie Loh (m.loh@imperial.ac.uk) Natalia Liem (csiv17@nus.edu.sg) Aparna Vaithilingam (csiav@nus.edu.sg) Pei Li Lim (csilimpl@nus.edu.sg) Nur Sabrina Sapari (csinurs@nus.edu.sg) Chee Leong Cheng (patv12@nus.edu.sg) Benedict Yan (benedict_yan@nuhs.edu.sg) Brendan Pang (brendan_pang@nuhs.edu.sg) Manuel Salto-Tellez (m.salto-tellez@qub.ac.uk) Wei Peng Yong (wei_peng_yong@nuhs.edu.sg) Barry Iacopetta (barry.iacopetta@uwa.edu.au) Richie Soong (csirs@nus.edu.sg) Version:2Date:23 October 2013 Author's response to reviews: see over
Cancer Science Institute of Singapore Dr. Jan Bornschein Associate Editor BMC Gastroenterology Wednesday, October 23, 2013 Dear Dr. Bornschein, Thank you very much for considering our manuscript for publication in BMC Gastroenterology. We would also like to thank the reviewers for their valuable time, comments and suggestions. Following is our point-by-point response to the reviewers comments (indicated in italics). Corresponding amendments to the text are marked by red text and underlining in the manuscript and supplementary tables. We have prepared a response that we believe will adequately address reviewers comments, given the many favorable comments and the allotted timeline for response. We would be willing to conduct further experiments should they be deemed necessary, although we would require a significant extension of response time to do so. We look forward to your kind consideration of our revision. Yours sincerely, Richie Soong Senior Principal Investigator, Cancer Science Institute of Singapore Research Associate Professor, Department of Pathology National University of Singapore Centre for Translational Medicine 14 Medical Drive, #11-02 Singapore 117599 Phone: +65 65168055 Fax: +65 68739664 e-mail: csirs@nus.edu.sg Centre for Translational Medicine, 14 Medical Drive, Singapore 117599 Tel: +65 65168055 Fax: +65 68739664 Website: www.csi.nus.edu.sg Company Registration No: 200604346E
REVIEWER 1 A paper submitted by Loh et al. represents an original research, which adds valuable information on methylation puzzle in gastric cancer. The authors use novel high throughput technologies, which enabled to assess methylation status in more than 1400 CpG sites. The methylation array used by the authors enables to examine multiple methylation sites, rather than looking at single CpG island. The authors describe two distinct methylation profiles in gastric cancer, which is of interest for further research. The authors discuss their findings with appropriate caution, reflecting on potential limitations and biases. The article is written in comprehensible English. Statistical analysis is appropriate for methods used within the study. I think that the article has adequate novelty and original data; therefore, I would suggest accepting the paper for publishing. Our response: We thank the reviewer for their kind comments. REVIEWER 2 The manuscript is well-written, and has very interesting findings about the CpG Island Methylator Phenotype (CIMP) in gastric cancer. Even though, the manuscript has some issues that should be clarified before its acceptance. Minor Revisions Page 3, Line 37: substitute the in for an Our response: The substitution has been made. Page 14, Line 267: H3K4-K3K27: is this right? If so, please explain what are the consequences of this configuration; Our response: There is no clear consequence of the H3K4-K3K27- histone marks configuration. We have therefore deleted the sentence [ The proportion of genes with H3K4-K3K27- marks was significantly lower in Group HG (4%) compared to Group LG (25%, P<0.001) and Group NG (24%, P<0.001). ] Page 17, Line 330: If only one CpG site was evaluated for most genes, what's the strength of the study? What are the bases to define which CpG site will be evaluated in each gene and its importance for the gene expression regulation? Our response: The strength of the study lies in the number of genes analyzed and the underlying design of the Goldengate array that focuses on cancer-related genes. The selection of which CpG site to evaluate in each gene is decided by the manufacturer (Illumina) as this is a pre-fabricated array (not custom-made). According to Illumina, these genes were selected based on their biological relevance in cancer, including tumor suppressor genes and oncogenes, and genes that are involved in cancer development. We have also obtained an update of the array content, and amended the text Moreover, only one CpG site was evaluated for most (70%) of these genes to Moreover, only two or less CpG sites were evaluated for most (86%) of these genes. Page 18, Lines 363-364: As the authors did not find genes differentially methylated between HP+ and HP- patients, explain if this could be a limitation of the technique. The same question may be applied for other clinicopathological features, such as TNM and histological type.
Our response: As discussed in the text (lines 369-390), this may due to the use of continuous versus binary values for methylation and to the thresholds used for statistical testing. It is also worth noting that as mentioned in the same section (lines 360-367), although we did not find any genes that were differentially methylated between HP- and HP+ GC tissue, we did find 8 genes that were hypomethylated in the normal gastric tissue of HP+ GC patients, including genes (DAB2IP and TWIST1) that have been implicated in gastric tumorigenesis, and STAT5A which was also previously observed to be hypomethylated in HP+ compared to HP- tissue from non-gc subjects, mirroring the current results observed in GC subjects. REVIEWER 3 In this study, Loh et al have investigated the methylation status of 60 gastric tumors and paired adjacent tissues using the Illumina GoldenGate Methylation Panel I assay which interrogates 1,421 CpG sites located within 768 cancer-related genes. The result showed that 219 CpG sites in 147 genes were differentially methylated between tumors and matched normal tissues, with HOXA5 and hedgehog signaling being the top-ranked gene and signaling pathway, respectively. Unsupervised clustering and statistical analysis indicated high and low methylation subtypes of gastric cancers, in which female patients were preferentially associated with the former subtype. No significant difference was found in other clinicopathological parameters such as age, metastasis, survival, H. pylori infection. Major Compulsory Revisions 1. Upon the genome-wide DNA methylation analysis and statistical analysis, the authors identified a number of potentially differentially-methylated genes between gastric tumor and non-tumor samples. However, as mentioned by the authors in Discussion, only one CpG site was evaluated for 70% of these genes. It is therefore important to validate these candidates using an independent and preferentially quantitative DNA methylation assay e.g. pyrosequencing or MethylCap assay. For example, whether the differential methylation of HOXA5 and the 8 genes in the hedgehog signaling pathway (Table 1) can be confirmed? Our response: As mentioned in our response to Reviewer 2, we have obtained an update of the information on array content, and amended Moreover, only one CpG site was evaluated for most (70%) of these genes to Moreover, only two or less CpG sites were evaluated for most (86%) of these genes (line 328). The correlation between GoldenGate and other recognized methylation platforms, such as pyrosequencing, has been previously well established in many studies (e.g. Christensen BC et al., PloS Genetics 2009; Amatruda JF et al., BMC Cancer 2013; Carmona FJ et al., Cancer Prev Res 2013). The further validation of the GoldenGate platform for 9 genes is an extensive effort that would extend the scope of this study beyond its initial goals of reporting the findings of high-throughput analysis of well-annotated samples. It also would exhaust scarce clinical study material, and prevent a timely response. We humbly request the reviewer s understanding in appreciating the results of this study for what they are. 2. It is now recognized that field effect of DNA methylation occurs in the gastric tissues adjacent to the tumor, even though they look histologically normal. In this study, the authors used tumor-adjacent tissues from the same patients for the DNA methylation analysis and claimed that these are normal gastric tissue (line 204 of page 11; line 224 of page 12; line 364 of page 18). Gastric tissues from non-cancer/inflammed patients should be as normal control. Our response: We definitely agree with the reviewer on the importance of field effect and is also an area of interest for the authors (Subramaniam MM, Loh M, et al., Mol Carcinog. 2012 Aug 21). However, it is the intention of the current manuscript to compare tumor and normal gastric tissue from
the same subject. Using gastric tissues from non-cancer/inflammed patients as normal controls will be testing a different hypothesis. 3. The authors described in the abstract and the main text that the existence of a comparable CIMP subtype in gastric cancer has not been clearly established because of relatively low interrogated CpG sites and sample size. However, as mentioned in the Discussion by the authors, Kim et al (Cancer Lett 2013) and Zouridis et al (Sci Transl Med 2012) have carried out more comprehensive studies in terms of much more interrogated CpG sites (over 27,000) and larger sample number (over 200). Both studies have revealed the existence of CIMP in subgroups of gastric cancers, which associated with distinct clinicopathological parameters. What additional information have the authors gained in this study? The new data and insight should be clearly described in the Discussion in comparison with these two studies. Our response: One important advantage of our current study is the use of a well-annotated series of tumor samples with high tumor content (>70%), increasing the robustness and accuracy of the results presented here. As mentioned in the Introduction, we have previously demonstrated (Loh M et al., Diagn Mol Pathol 2010) that the use of samples with low tumor content significantly influences the interpretation of methylation levels and candidate gene identification. We acknowledge that the two studies by Kim et al and Zouridis et al are more comprehensive in terms of CpG sites and sample size, but we believe that the high-quality tissues used in our study allows us to add value to the available scientific knowledge too. 4. Some references are either mis-numbered (e.g. #16 in line 92; #8 in line 109; #28 in line 135; #32 in line 153) or missing (#45-56), which make the manuscript really hard to follow. Our response: We have corrected the referencing, which resulted from referencing inconsistency between co-author versions of the manuscript. We apologise for the inconvenience caused.