Application of the High-Throughput TAB-Array for the Discovery of Novel 5-Hydroxymethylcytosine Biomarkers in Pancreatic Ductal Adenocarcinoma

Round 1

Reviewer 1 Report

 

 

Dear Editor,

 

I have carefully reviewed the manuscript entitled “5-Hydroxymethylcytosines as Novel Epigenetic Biomarkers for Pancreatic Adenocarcinoma” by Zeng, et al. As it relates to the stringency of the study, I have evaluated the quality of the manuscript experimental design, presentation, writing, and illustration. I have also evaluated the amount of original information that the manuscript provides by extensively investigating the amount and type of information previously published in the field. Following, you will find what I believe is a thorough and fair assessment of the manuscript. I also provide detailed recommendations that the authors can use to improve their manuscript. Thus, I am optimistic that you and the authors will find these constructive comments useful.

 

 

A. GENERAL DESCRIPTION: The authors describe 5-hydroxymethylcytosines (5hmC) profiles in 17 pairs of pancreatic tumor/adjacent tissue samples, using the TAB-Array assay, which is a combined technology to directly profile 5hmC at single base resolution with the Illumina EPIC array through TET-Assisted Bisulfite (TAB) conversion. The study finds distinct distribution patterns of 5hmC in tissue samples from PDAC patients relative to gene regulatory elements, in particular significant enrichment in enhancers (marked by H3K4me1), regions with active transcription (marked by H3K27ac), gene bodies (marked by H3K36me3), and a depletion at promoters (marked by H3K4me3), regions with polycomb repression (H3K27me3), and formation of heterochromatin (H3K9me3), as identified by the Roadmap Epigenomics Consortium. The authors identify differences in 5hmC modified CpG sites, involving genes related to cancer pathobiology, between tumor and adjacent tissues, some of which also showed prognostic value for PDAC patient survival based on data from the Cancer Genome Atlas Project. Thus, the authors propose that these studies establish the potential of 5hmC as a novel epigenetic biomarker for PDAC diagnosis and prognosis.

 

B. STRENGTHS: The authors look at the relatively newly studied epigenetic mark 5hmC, which is gaining interest regarding its role in the pathobiology of cancer and its potential as a biomarker.  The ability to evaluate this epigenetic mark in clinical samples has been fairly limited due to the technologies to distinguish this mark from other 5mC derivatives. While prior techniques to separate cytosine modifications were solely sequenced based, recently, new techniques have been combined with the more affordable Illumina EPIC array through TET-Assisted Bisulfite (TAB) conversion, which was employed in the current study. Previous reports have shown differences in 5hmC in PDAC using PDAC cell lines and patient-derived xenografts through the use of these sequencing methodologies, however this is the first study this reviewer is aware of using the TAB-array methodology. Thus, the authors should be commended for their efforts.

 

C. SUGGESTIONS FOR IMPROVEMENT:

 

 

1.    In the discussion, the authors admittedly recognize the limitations of the study in regard to biomarker potential, which the reviewer completely agrees with these reflections. As a result, it seems that the title of the manuscript and other statements in the abstract (lines 30-31), are overstated. The study has value in that it uses an emerging technology that can be readily adapted to initiate additional studies in this field to truly establish 5-hmC-based biomarkers. Therefore, the authors should be careful in what they state the study “establishes” versus the true value of the study, which is present and needs to be clarified. Otherwise, additional experimentally-based evidence would be required rather than predictions. Focusing on the utility of 5-hmC as a biomarker (for example extensively in the objective of the study and introduction), while a valid discussion point, is not necessarily strengthening the presentation of the paper. There are missed opportunities to highlight true strengths of the study.

2.    The authors do not give consideration to the minimum 2 widely accepted subtypes in the field (classical/basal, classical/quasi-mesenchymal, pancreatic progenitor/squamous, for which classical/basal classification is supported by TCGA: 10.1016/j.ccell.2017.07.007). It would be prudent for the authors to consider the 2 subtypes and comment on whether 5-hmC patterns were consistent across all tumors or if stratification into subgroups occurred.

3.    While the authors clarify that “down-regulated 5hmC loci” means “(i.e., lower modification levels in tumor samples)” (lines 122-123), this wording can be confusing to the reader and should just be stated as it is – lower 5hmC levels rather than requiring a clarification in parenthesis. 

4.    Details regarding the methods for measuring the enrichment of tumor-associated 5hmC loci with various genomic features (e,g. histone modification marks) is vague. The presence of the histone marks versus the 5-hmC is not detailed for location relative to each other (within a particular distance or overlapping?), cut-off of signal for the histone marks to indicate presence, etc. The authors should provide clarification of the methods.

5.    As referenced by the authors, several studies have been performed for expression, DNA methylation and histone marks in PDAC with patient samples, which are publicly available datasets. The current study would be significantly strengthened with some integration of these existing studies as the pancreas data from the Roadmap Epigenomics Project (Nature, 2015) is based on one normal whole pancreas sample.

Author Response

The reviewer’s comments are italicized.

 

Point 1: In the discussion, the authors admittedly recognize the limitations of the study in regard to biomarker potential, which the reviewer completely agrees with these reflections. As a result, it seems that the title of the manuscript and other statements in the abstract (lines 30-31), are overstated. The study has value in that it uses an emerging technology that can be readily adapted to initiate additional studies in this field to truly establish 5-hmC-based biomarkers. Therefore, the authors should be careful in what they state the study “establishes” versus the true value of the study, which is present and needs to be clarified. Otherwise, additional experimentally-based evidence would be required rather than predictions. Focusing on the utility of 5-hmC as a biomarker (for example extensively in the objective of the study and introduction), while a valid discussion point, is not necessarily strengthening the presentation of the paper. There are missed opportunities to highlight true strengths of the study.




Response: We agreed with the reviewer that the true value of the current study is in the technological foundation for future studies by demonstrating the feasibility of applying the TAB-Array in cancer biomarker discovery. Specifically, our findings also suggested the potential of 5hmC as novel epigenetic biomarkers for PDAC diagnosis and prognosis. As suggested, we toned down the statements in the title and abstract, and discussed the potential of TAB-array in the revised text to highlight the true strengths of this study (See Page 2 Line 69-73; Page 6 Line 185-187; Page 7 Line 201-202). We also changed “establish” to “suggest” or “demonstrate” throughout the text to be more accurate.

 

Point 2: The authors do not give consideration to the minimum 2 widely accepted subtypes in the field (classical/basal, classical/quasi-mesenchymal, pancreatic progenitor/squamous, for which classical/basal classification is supported by TCGA: 10.1016/j.ccell.2017.07.007). It would be prudent for the authors to consider the 2 subtypes and comment on whether 5-hmC patterns were consistent across all tumors or if stratification into subgroups occurred.

 

Response: We look forward to integrating 5hmC patterns with other genomic data such as mRNA and lncRNA to provide a more comprehensive picture of the relationship between 5hmC profiles in cfDNA and the molecular characteristics of PDAC. However, we were not able to perform this analysis due to the unavailability of transcriptomic data and subtype information of the analyzed tumour samples. We commented this point in discussion as part of future directions (Page 6 Line 196-199), though the primary aim of the current study was just to explore the biomarker potential of 5hmC in PDAC using a novel technique.

 

Point 3: While the authors clarify that “down-regulated 5hmC loci” means “(i.e., lower modification levels in tumor samples)” (lines 122-123), this wording can be confusing to the reader and should just be stated as it is – lower 5hmC levels rather than requiring a clarification in parenthesis.

 

Response: We re-wrote this sentence and changed similar statements throughout the text to provide a clear description. Please see Page 4 Line 125, 127-128, 136.

 

Point 4: Details regarding the methods for measuring the enrichment of tumor-associated 5hmC loci with various genomic features (e,g. histone modification marks) is vague. The presence of the histone marks versus the 5-hmC is not detailed for location relative to each other (within a particular distance or overlapping?), cut-off of signal for the histone marks to indicate presence, etc. The authors should provide clarification of the methods.

 

Response: In this study, the presence of pancreas-derived histone marks was determined based on the Gapped Peaks (i.e., narrow contiguous regions of enrichment for histone ChIP-seq and DNase-seq) identified in the Roadmap Epigenomics Project. According to the documentation from Roadmap, those narrow peaks that passed a Poisson p value of 0.01 were defined as histone modification regions. Specifically, we retrieved the start and end annotations of the Gapped Peaks from the Roadmap. The 5hmC loci detected by the TAB-Array that overlapped with these Gapped Peaks were considered to be co-localized with histone marks and were included for further statistical analysis. As suggested, we provided more details in “4.4. Evaluating genomic distributions of 5hmC and functional relevance”. Please see Page 8 Line 257 - 260.   

 

 

Point 5: As referenced by the authors, several studies have been performed for expression, DNA methylation and histone marks in PDAC with patient samples, which are publicly available datasets. The current study would be significantly strengthened with some integration of these existing studies as the pancreas data from the Roadmap Epigenomics Project (Nature, 2015) is based on one normal whole pancreas sample.

 

Response: We agreed with the reviewer that incorporating mRNA expression, DNA methylation and ChIP-Seq data derived from PDAC patient samples would strengthen the current study. In the current study, we utilized the histone modification data derived from normal pancreas to demonstrate the enrichment of tumor-associated 5hmC loci (when compared with tumor adjacent tissue) with pancreas-derived histone marks. In addition, we obtained the transcriptome data for PDAC from The Human Pathology Atlas (HPA) to demonstrate the prognostic potential of 5hmC signatures for PDAC. In this revision, we obtained the pancreatic cancer-specific mRNA expression data from the HPA, a permutation test (N = 100,000) on the 822,760 CpGs sites indicated enrichment of tumor-associated 5hmC loci (when compared with tumor adjacent tissue) with pancreatic cancer-specific expression signatures (empirical p-value = 0.004). We provided more details in “4.5. Exploring prognostic value and pancreatic cancer-specific expression of PDAC-associated 5hmC” & “2.3. Exploring Functional Relevance, Prognostic Value and Cancer Specificity”. Please see Page 4-5 Line 139 – 143 & Page 8 Line 129-132 & 138-142. However, since previous studies of pancreatic cancer epigenetics, including TCGA 450k array data did not distinguish 5hmC from 5mC, a direct comparison with existing methylation results would be difficult to interpret at this moment. Our findings suggested the importance of differentiating between 5mC and 5hmC in future studies using emerging technologies such as the TAB-Array.

 

Reviewer 2 Report

The work of Chang Zheng et al explores the distribution of the epigenetic marker 5-hydroxymethylcytosine (5hmC) in a set of pancreatic carcinoma samples in comparison with healthy tissues.

The work is essentially a validation of the Illumina TAB-Array method for analysis of differential 5hmC modification in patients' samples. The work is well-written, technical sound and accurate and conclusions are supported by experimental work. However, the contribution of the work to knowledge advancement is very limited, mainly due to lack of experimental validation of promising target genes.

Some revision may be done to enhance the strenght of the work.

Major issues:

As already stated, a validation of at least some target genes would be appropriate. e.g. Measure expression levels of target genes with differential 5hmC in tumor samples compared to non-tumor. If this is not possible due to unavailability of patients' samples, the authors may rafer to public databases (e.g. the HPA database they use in paragraph 2.3), not only to evaluate association with prognosis but also to evaluate association of altered 5hmC profile with altered expression level. This would add information on the biological role of 5hmC on gene expression regulation in this setting.

Some information about clinical and pathological features of patients would be appropriate. Although the number of patients is very limited, the Authors could try to correlate 5hmC alteration on selected target genes at least with overall survival.

In paragraph 2.3 the Authors perform a KEGG and GO enrichment analysis on genes associated with differential 5hmC. It would be useful to distinguish between genes showing increased 5hmC levels and genes showing decreased 5hmC levels. Again, this would add information on the biological role of this modification.

Minor issues:

It is not clear weather the Authors started from FFPE tissues, fresh tissues or nitrogen-frozen tissue.

In Figure 2 legend, PDAX should be corrected to PDAC

Author Response

The reviewer’s comments are italicized.

 

Point 1: As already stated, a validation of at least some target genes would be appropriate. e.g. Measure expression levels of target genes with differential 5hmC in tumor samples compared to non-tumor. If this is not possible due to unavailability of patients' samples, the authors may refer to public databases (e.g. the HPA database they use in paragraph 2.3), not only to evaluate association with prognosis but also to evaluate association of altered 5hmC profile with altered expression level. This would add information on the biological role of 5hmC on gene expression regulation in this setting.

 

Response: As suggested, in this revision, we obtained from the HPA the gene expression data of TCGA PDAC samples and performed additional analyses. Specifically, our detected 5hmC markers for pancreatic cancer were found to be enriched with pancreatic cancer-specific genes from the HPA. We updated the text on Page 5 Line 138-142 and Fig. 3C.

 

Point 2: Some information about clinical and pathological features of patients would be appropriate. Although the number of patients is very limited, the Authors could try to correlate 5hmC alteration on selected target genes at least with overall survival.

 

Response: Our primary aim was to confirm that the high-throughput TAB-Array approach would be technically feasible for biomarker discovery in pancreatic cancer. We hope that findings from this study would support the use of this novel approach in future larger scale studies with more samples to investigate relevant clinical questions, including survival. Unfortunately, we do not have clinical outcomes for those patients whose tissue samples were included in this study. Therefore, instead of a direct analysis of survival, we chose to evaluate the enrichment of the 5hmC-contatining marker genes in the prognostic genes for pancreatic cancer from the HPA database. Interestingly, the current analysis suggested the prognostic potential of 5hmC modifications in pancreatic cancer based on the HPA and TCGA PDAC samples. We discussed this limitation and look forward to applying the TAB-Array in our future studies for more comprehensive analysis and further validation. Please see Page 6-7 Line 196-199.  

 

Point 3: In paragraph 2.3 the Authors perform a KEGG and GO enrichment analysis on genes associated with differential 5hmC. It would be useful to distinguish between genes showing increased 5hmC levels and genes showing decreased 5hmC levels. Again, this would add information on the biological role of this modification.


 

Response: As suggested, we updated the results from the KEGG and GO enrichment analysis in “2.3 Exploring Functional Relevance and Prognostic Value and Cancer Specificity”. Please see Page 4 Line 124-128. Specifically, 1048 out of 1118 PDAC-associated 5hmC loci were “down-modified” in tumor tissues and genes containing these differential 5hmC loci were enriched in such pathways as “PI3K-Akt signaling pathway”, “focal adhesion”, “Rap1 signaling pathway”. Interestingly, genes co-localized with “up-modified” 5hmC loci were enriched in “cholinergic synapse (GNAO1, GNG4, GNG7)”.

 

Point 4: It is not clear whether the Authors started from FFPE tissues, fresh tissues or nitrogen-frozen tissue.

 

Response: The genomic DNA was prepared from snap frozen tissues. We clarified this point in “4.2. DNA isolation and the TAB-Array assay”. Please see Page 7 Line 224.

 

Point 5: In Figure 2 legend, PDAX should be corrected to PDAC

 

Response: We corrected this typo. Please see Page 4 Line 112.

 

 

 

Round 2

Reviewer 1 Report

The authors have provided additional clarifications as requested. I have no further comments.

Reviewer 2 Report

The Manuscript has been improved. I have no further concerns