microRNA Profile Associated with Positive Lymph Node Metastasis in Early-Stage Cervical Cancer

Round 1

Reviewer 1 Report

In this study, authors found a subset of miRNAs that may serve as molecular markers for early-stage CC patients with lymph node metastasis, who may not be identified by imaging.

1. Is the HPV status of patients considered as a factor to match in the analysis?
2. Table 2 may be presented before Figures 2 and 3 for easier understanding of the content.
3. Table 3 is not presented but is cited in the main text.
4. For result section 3.5, what are the two % correspond to?

Author Response

Response to Reviewer 1 Comments

 

1. Is the HPV status of patients considered as a factor to match in the analysis?

 

Response 1:  As part of the initial approach for the diagnosis of cervical cancer, the different management guidelines do not recommend HPV status, since the management of these patients would not change, because of that, at this time We did not consider it unrewarding to delve further into this feature of cervical tumors and the analysis initially proposed.

 

2. Table 2 may be presented before Figures 2 and 3 for easier understanding of the content.

 

Response 2: Thank you for your comment, we have made the suggested changes. We have relocated figures 2 and 3 within the main text for a better understanding by the reader.

 

 

 

 

3. Table 3 is not presented but is cited in the main text.

 

Response 3: Thank you for pointing this out. We did not include table 3 by mistake. In this new version, table 3 is now referred as a table 2 and it is located between lines 259 to 270.

 

4. For result section 3.5, what are the two % correspond to?

 

Response 4: It intended to show the percentage of patients who expressed some specific miRNA above the mean; however, as it also was suggested by the other reviewer, it was changed to median with interquartile range.

Reviewer 2 Report

The manuscript by Abraham-Muñoz and coworkers describes a miRNA-based signature for the potential detection of lymph node metastasis in early-stage cervical tumors. The paper is well written and structured, and the authors took a lot of care in describing the experimental protocols, which must be considered as a very positive fact. I would like to personally congratulate the authors, because this is not very common in recent times, and I appreciate the scientific soundness of the methods.

Despite the small number of patients recruited for the study, the results are solid and well founded in the obtained data. I would definitively recommend the article for publication after the authors consider my suggestions.

 

Point-to-point comments

1.- Figure 1: legend should be completed indicating the algorithm used for sample clustering. Also within this context, is anything special with the two samples that are not clustered in their proper groups?

2.- Figures 2,3 and table 2 would need to be reformulated. I would suggest the authors to graphically represent only the statistically significant differentially expressed miRNAs in just one figure. On the other hand the table should contain all the data including significant and non-significant values. In the graphs please use a graphical nomenclature to show the p-values (for instance asterisks)

3.- I understand that the microarrays used by the authors to perform the screening could have an outdated miRNA nomenclature, but I would also recommend that in the whole paper the authors used the more updated miRNA nomenclature based on miRbase 22.1. This should contain the prefix for the species and the corresponding suffixes for 5p- or 3p- versions of the miRNAs. Please avoid to use the denomination “star” since this is not used anymore.

4.- The authors should strength the quality of the obtained data and information by a careful functional analysis of the selected miRNAs. I would recommend showing graphically the results of the regulatory networks established by the selected miRNAs. If the authors proposed to use some of them as biomarkers of the development of metastasis, it would be very interesting for the readers to know which are the pathways and genes regulated by those miRNAs. There is a reference of these results in Table 3, but I was not able to find it within the text. Please correct this accordingly and try to produce some graphical results like bar graphs for pathway or GO-term enrichment terms. Be careful and divide the selected miRNAs in two groups considering the UP and DOWN regulated miRNAs separately for GO and pathway analysis.

Author Response

Response to Reviewer 2 Comments

 

1. Figure 1: legend should be completed indicating the algorithm used for sample clustering. Also within this context, is anything special with the two samples that are not clustered in their proper groups?

 

Response 1: Thank you, as you had suggested, we have included this information in the materials and methods section, and can be found on lines 124-126. Additionally, we have added more information about it in the legend of Figure 1.

 

2. Figures 2,3 and table 2 would need to be reformulated. I would suggest the authors to graphically represent only the statistically significant differentially expressed miRNAs in just one figure. On the other hand the table should contain all the data including significant and non-significant values. In the graphs please use a graphical nomenclature to show the p-values (for instance asterisks)

 

Response 2: We thank you for the comment and We agree. We have merged Figures 1 and 2 into a single figure and presented only the miRNAs that were significant in the validation experiments according to your suggestion (you can see this in figure 2 in this new version). Similarly, we have indicated in each of the figures the p value for each experiment, we believe that indicating the corresponding p value in each experiment is more informative than adding graphical nomenclature to show the p-values (for instance asterisks).

 

3. I understand that the microarrays used by the authors to perform the screening could have an outdated miRNA nomenclature, but I would also recommend that in the whole paper the authors used the more updated miRNA nomenclature based on miRbase 22.1. This should contain the prefix for the species and the corresponding suffixes for 5p- or 3p- versions of the miRNAs. Please avoid to use the denomination “star” since this is not used anymore.

 

 

Response 3: Thank you, we have updated the nomenclature of all the miRNAs used in this work both in the main text and in the figures and tables included in this work.

 

4. The authors should strength the quality of the obtained data and information by a careful functional analysis of the selected miRNAs. I would recommend showing graphically the results of the regulatory networks established by the selected miRNAs. If the authors proposed to use some of them as biomarkers of the development of metastasis, it would be very interesting for the readers to know which are the pathways and genes regulated by those miRNAs. There is a reference of these results in Table 3, but I was not able to find it within the text. Please correct this accordingly and try to produce some graphical results like bar graphs for pathway or GO-term enrichment terms. Be careful and divide the selected miRNAs in two groups considering the UP and DOWN regulated miRNAs separately for GO and pathway analysis.

 

Response 4: Thank you for pointing this out, this suggestion has been very significant in order to increase the impact of our work. First, we have added Table 2 (previously referred to as Table 3), which contains the biological processes enriched by the target genes (GO-terms) as well as the genes included in each process. In addition, we have constructed the proposed interaction network between the target genes and the 7-miRNA signature (Figure 3).

 

 

Author Response File: Author Response.doc

Reviewer 3 Report

Review

microRNA profile associated with positive lymph node metastasis in early-stage cervical cancer

 

Thanks for the potentially interesting paper on the identification of de-regulated microRNAs for lymph node positive cervical cancer. There are some technical issues, which hampers the evaluation of the results.

 

The results of the microarray experiment should be corrected for multiple testing. If this is not done the number of false positives are too high for the results to be trusted.

 

Figure 1: It is not clear what values are plotted in figure 1 and how the data is processed. 

It could be relevant to add the most important clinical parameters in the heatmap. Increase font size. Remove the hsa-. I am curios if anything interesting distinguish the two samples from the other samples, which cluster in with the “wrong” set of samples.

 

Figure 2: keep all the plots together in one and make a correlation plot of the two platforms. The color labels has been swapped in these plots. So the non-LM is red. It is not clear to me what the plotted values are? If these are delta Ct values, the differences are very small and can not be considered important. Increase font sizes

 

Are the values used in table 2 the qPCR data or the array data? Don’t use asteric in the table. Asteric has previously been used to denote microRNA major and minor variants and could therefore confuse the readers. Better to split the table in two parts, one for the upregulated and one for the downregulated.

 

Why not do a logistic regression between the expression values and the lymph node status? The arbitrary split on expression values, does not make sense to me.

 

I would suggest the authors to start with an unsupervised analysis of the samples to investigate for possible patterns in the data. The squamous carcinomas normally have a distinct expression pattern of microRNAs and might not be comparable to the adenocarcinomas. So the 3 adenosquamous could be taken out of the initial analysis if considered outliers.

 

Reference on the delta delta Ct method

 

Include the expression data as supplementary material and/or in GEO database.

 

Line 244-245:  Is the overlap of 5 microRNAs between the two sets significant? What is meant by the last part of the sentence? (“ and several gene targets that are associated with LNM+ samples”)

 

How does the authors plan to use the 7-miR signature? The cutoff values are based on the mean expression in a small cohort of samples. These values are not easily transferred to a new set of samples. An idea is to build ratios of the differentially expressed microRNAs and used these in a classifier.

Author Response

Response to Reviewer 3 Comments

 

1. Figure 1: It is not clear what values are plotted in figure 1 and how the data is processed. It could be relevant to add the most important clinical parameters in the heatmap. Increase font size. Remove the hsa-. I am curios if anything interesting distinguish the two samples from the other samples, which cluster in with the “wrong” set of samples.

 

Response 1: Thank you for your valuable comment. We have added important information to the legend of Figure 1 to clarify the analytic methodology and the plotted data. Likewise, we have added detailed information in the materials and methods section related to the elaboration of this figure, this information can be consulted in lines 124-126. In addition, we have carefully reviewed the clinical and pathological information of samples NLM013 and LM011 and found no notable clinical, pathological, or molecular differences with respect to the rest of the samples in the study group. We cannot explain why these samples are grouped in a different group with respect to their study group, however we believe that excluding these from the study with the intention of improving our detection rates is not correct and would be disingenuous. For this reason, we have decided to set out these results without any modifications.

 

 

 

 

 

2. Figure 2: keep all the plots together in one and make a correlation plot of the two platforms. The color labels has been swapped in these plots. So the non-LM is red. It is not clear to me what the plotted values are? If these are delta Ct values, the differences are very small and can not be considered important. Increase font sizes

 

Response 2: We thank you for the comment. We have created a single figure from the results of the validation experiments (RT-qPCR). We apologize for the confusion. The colors used in Figure 2 are related to the expression value of each miRNA according to the key color presented in the heatmap included in this new version (blue for overexpression and red for underexpression, z-score). Thus, in the plots of Figure 2, the relative overexpressed values are plotted in blue for both sample groups (LM and NLM) while the relative underexpressed values are plotted in red for both sample groups as plotted on the heatmap (results from microarray experiments).

 

3. Why not do a logistic regression between the expression values and the lymph node status? The arbitrary split on expression values, does not make sense to me.

 

Response 3: The suggested change was made. A logistic regression was performed taking into account the expression of the miRNAs and the lymph node status; likewise, the values ​​in table 3 (previously table 2) and in the manuscript were changed according to your suggestion and comment.

 

 

4. I would suggest the authors to start with an unsupervised analysis of the samples to investigate for possible patterns in the data. The squamous carcinomas normally have a distinct expression pattern of microRNAs and might not be comparable to the adenocarcinomas. So the 3 adenosquamous could be taken out of the initial analysis if considered outliers.

 

Response 4: The unsupervised analysis was not performed because the objective of the study is to identify a profile of miRNAs associated with lymph node involvement. On the other hand, in the comparative analysis of the histopathological variables, no statistical difference was identified between different histologies (p = 0.57). Finally, due to the small number of adenocarcinoma and adenosquamous histologies, we believe that the results of unsupervised analysis may not be representative.

 

 

5. Reference on the delta delta Ct method

 

Response 5: The following reference was added: Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. . 2001 Dec;25(4):402-8.  doi: 10.1006/meth.2001.1262. In lines 137-138

 

6. Include the expression data as supplementary material and/or in GEO database.

 

Response 6: The expression of the requested microarray experiments is attached in the supplementary material, as requested

 

7. Line 244-245:  Is the overlap of 5 microRNAs between the two sets significant? What is meant by the last part of the sentence? (“ and several gene targets that are associated with LNM+ samples”)

 

Response 7: No statistical analysis was performed between the two sets. It was only a review of the literature where 5 of our miRNAs coincided with what was described in the authors Chen et. al. We change the sentence for better understanding: “Interestingly, our results are consistent for 5 miRNAs (miR-548, miR-379, miR-337, miR-487b and miR-376c) and several gene targets associated with lymph node involvement.” Now, this sentence is in line 283.

 

 

8. How does the authors plan to use the 7-miR signature? The cutoff values are based on the mean expression in a small cohort of samples. These values are not easily transferred to a new set of samples. An idea is to build ratios of the differentially expressed microRNAs and used these in a classifier.

 

Response 8: The suggestion was made to take into account miRNA expression instead of median cutoff to make it easier to transfer to a new set of samples for future studies.

Author Response File: Author Response.doc

Round 2

Reviewer 3 Report

Thanks for the resubmission.

 

There is something wrong with the qRT-PCR values shown in the paper. The values are very low. If the authors calulate delta Ct values and get eg. miR-92b-5p to be differentially expressed in the order of 0.005. This would correspond to a difference between the mean of sample cohorts of less than 1%. Please provide the raw data and/or redo the analysis to show difference closer to the expected order of magnitude. Eg. miR-92b-5p is expressed (0.87 log2, 1.82 fold change 82% higher in LM group).

 

Do a correlation plot of the qRT-PCR data and the microarray data.

 

Provide the qRT-PCR data as supplements.

 

Provide supplements of the target prediction analysis.

 

Increase the fonts of Figure 1. I would use row mean centered data instead of z-scores can distort small changes for individual microRNAs.

 

Provide all the microarray data as supplements.

 

Make a table of the microarray significant analysis and include it in the supplements.

 

Author Response

Here, it is attached the POINT-BY-POINT response to the reviewer.

 

Thank you for your valuable comments, we appreciate the time and effort that you have dedicated to providing your valuable feedback on this work. We have made several modifications according to the suggestions.

 

 

Response to Reviewer 3 Comments Round 2

 

 

Here, we list each of the comments to answer them:

 

1.- There is something wrong with the qRT-PCR values shown in the paper. The values are very low. If the authors calulate delta Ct values and get eg. miR-92b-5p to be differentially expressed in the order of 0.005. This would correspond to a difference between the mean of sample cohorts of less than 1%. Please provide the raw data and/or redo the analysis to show difference closer to the expected order of magnitude. Eg. miR-92b-5p is expressed (0.87 log2, 1.82 fold change 82% higher in LM group).

 

Response 1:  Thank you for your comments, them are of the utmost importance to us, we hope that with these data the doubts will be resolved. We have carefully reviewed the analysis performed on the data obtained from the validation experiments using Taqman probes. First, we selected 10 miRNAs to validate their expression levels according with the microarray results, 5 over-expressed miRNAs and 5 under-expressed miRNAs. To perform the qPCR analysis, we used RNU6B expression (assay ID: 001093) as an endogenous control and carefully followed the reported method to calculate the relative expression with respect to a reference gene (RNU6B) in both sets of samples. Then, we used the 2^Ct values to elaborate the presented graphs (Figure 2). This information was specified in line 134 of this new version of the manuscript. We decided to use this method and not the Fold Change to show the differences in each sample within the analyzed groups (LM vs NML). Finally, we conducted a t-student analysis to evaluate the differences between the two groups.

 

Although the differences in magnitude between the two groups are small, they are sufficient to demonstrate a statistically significant correlation and corroborate the results obtained in the microarray experiments. These types of differences between expression levels have also been reported in other papers, for example in PMID: 22099053, PMID: 32078636 and PMID: 30562933. We believe it is desirable that in these experiments the differences between expression levels are much wider, however, in our experiments the differences are more discrete but sufficient to demonstrate a correlation with the microarray experiments.

 

 

2.- Provide the qRT-PCR data as supplements.

Response 2:  According to your suggestion, we will make qPCR data available upon request. It is attached as Supplement3.xlsx and Supplement2.cvs

 

3.- Provide supplements of the target prediction analysis.

Response 3:  We have added supplementary information on the target prediction analysis in the supplementary file 3 (line 217 on the manuscript)

 

4.- Increase the fonts of Figure 1. I would use row mean centered data instead of z-scores can distort small changes for individual microRNAs.

Response 4: We have enlarged the lettering in figure 1 as much as possible without distorting the figure, so as not to lose the proportions in the names.

 

5 .-Provide all the microarray data as supplements. Make a table of the microarray significant analysis and include it in the supplements.

 

Response 5: According to your suggestions we have made all data concerning microarray experiments available under a simple request to the corresponding author, line 367 (Data Availability Statement).

 

 

We hope that with these changes we significantly cover the concerns and comments of the reviewer and make the document suitable for publication in Current Oncology Journal.

 

Very Kind Regards,

 

 

 

David F. Cantu-de León

Division of Research

Instituto Nacional de Cancerología

Ciudad de México, Tlalpan ZP. 14080

dcantude@gmail.com

 

Round 3

Reviewer 3 Report

Thanks for the reply.

I will comment on the response in the same numerical order as the point-by-point response of the second review. I hope the authors claim of

 

1) The authors fail to answer the question. The difference in the array is 82% higher in the LM group and in the qRT-PCR data the difference is less than 1%. This is simply not a validation of the microRNA array data. The qRT-PCR data should in principle be corrected for multiple testing eg. divide the p-value by 10 (as the authors did 10 different tests). Then the “validation” will fail for miR-4534, miR-483-5p, miR-487b, miR-195 and miR-29b-2-5p. There is therefore only statistical support for miR-548ac and borderline support for miR-92b-5p. Which makes a “validation” of 2/10 microRNAs. This is not impressive. The papers referenced:

PMID: 22099053 – figure 3 (this is data on the log10 scale, so the differences are much larger than presented in this paper.)

PMID: 32078636 – figure 4 (miR-31-3p has the same order of difference as the manuscript, but are not significant, the other microRNAs are much order of difference larger)

PMID: 30562933 – I cannot find any data in this paper to support the author's claim.

I think the authors should evaluate if they trust the qRT-PCR data or the array more?

There seems to be a very large discrepancy between the two platforms. The differences in the array experiment should be able to be found in the qRT-PCR at a higher rate than seen here. Is RNU6 the best normalizer? Test other microRNAs from the array with relatively high expression and stable expression across samples. Try other qRT-PCR methods?

 

Another note for the authors could be to look at the primary transcripts of the microRNAs. Eg. miR-92b-5p is normally the lesser expressed microRNA of the pri-mir-92b. what are the levels of miR-92b-3p? and is this differentially expressed between sample groups? Array data are more sensitive to cross detection of longer RNA species, so more cross detection to the pre-mir can be expected.

 

2) If the authors wish to resubmit, I would like to see the data and the calculations. (supplement3.xlsx is the target prediction analysis and supplement2.csv is the kegg term significance table). Please stick to the same format of tables when submitting. I prefer to have one xlsx file with multiple tabs, with one tab with descriptions of the other supplementary files.

3) Describe the headers of the supplementary file.

4) The height of the heatmap can be increased to make the font size larger.

5) ok

 

Author Response

We have tryed to complete the document with the changes requested, 

Author Response File: Author Response.pdf

Round 4

Reviewer 3 Report

Thanks for the effort to supply the information needed to evaluate the paper.

 

• After revision of the excel sheet, I have found the mistake. The calculations for ddct uses the wrong field e.g. field N4 should be subtracting L4 from J4. And in AE4 = AA4-AC4. The error is in all the ddct calculations. If you correct these errors the percentage difference fits better with the array.
• Clarify in the paper that there are only used a subset of the original samples for RT-qPCR validation. Match the names of the samples to the samples of the array.
• Round values to meaningful digits in the supplements.

Author Response

.

Author Response File: Author Response.pdf