Reviewer's report 1
Dr. Pierre Pontarotti, Université d'Aix Marseille, Marseille, France
The authors perform a retrospective analysis showing that Diffuse Large B cell lymphoma patients with elevated expression of protein Kinase C-Beta gene have worse prognosis.
This is an important confirmative result that should be published.
I like very much the idea of performing retrospective analysis using available data.
I recommend the authors to make a brief comment on the state of the art on such approaches and especially in the case of microarray gene expression databases. I think that this kind of approach will be more and more used in the future.
We appreciate the comments. Particularly, we agree with the reviewer that retrospective analysis of available microarray gene expression data is a "state of art" approach in biomedical research to generate novel findings by using the existing data. As recommended by the reviewer, we added several sentences in the 1st paragraph of the discussion to emphasize this point.
Reviewer's report 2
Dr. Kateryna Makova, Pennsylvania State University, Pennsylvania, United States
In this paper the authors show that increased PKC-b expression in some types of cancers can be correlated with a shorter survival time after treatment.
Authors should highlight the novelty of this study better.
We followed the reviewer's advice and highlighted the novelty in both abstract and the discussion section. For example, we added the statement "We present a first detailed pharmacogenomics report comparing PKC-β mRNA expression across different lymphoid malignancies and evaluating it as an outcome predictor" in the abstract.
1. In general this section could be expanded, in particular section on gene expression analysis that will be of interest to genomicists.
We agree with reviewer's opinion. Please see below our responses to each of reviewer's specific suggestions.
2. Readers may be unfamiliar with the lymphoma subtypes, and so a description of each would be beneficial.
We followed the reviewer's advice and added a paragraph at the beginning of the result section to briefly describe the lymphoma subtypes.
3. When comparing differentially expressed clones between quartiles 1 and 4 of PKC-b expression, it would be useful to compare both quartiles to healthy lymphatic tissues as a control.
We completely agree with reviewer's opinion. However, the Rosenwald dataset that we analyzed did not include healthy lymphatic tissues as a control. Therefore, we could not compare PKC-β expression in DLBCL with that of healthy control tissues.
4. There are now protein-protein interaction studies that can be included in the study to identify the protein partners of the expressed mRNA.
We agree with the reviewer that protein-protein interaction data could provide more insights into the role of PKC-β in DLBCL. However, the main goal of the current study was to evaluate PKC-β as a target of B lymphocyte malignancies by analyzing its expression across subtypes and correlation between expression and survival, and our results have clearly indicated that PKC-β could be an important target for B cell lymphoma chemotherapy. Protein-protein interaction would exceed the scope of this study.
5. What genes are under-expressed in the quartiles, compared to each other, and compared to healthy tissues?
We followed the reviewer's suggestion and analyzed genes under-expressed in quartile 4(high PKC-β expressors). Unlike those over-expressed genes, we did not observe specific pathways with significantly under-expressed genes correlated with high PKC-β expression.
1. It is a little confusing that the comparison of PKC-b expression across different kinds of cancers is included at the beginning of the Discussion section. This seems to justify further studies of PKC-b in CLL and FL cancers because it was shown to have potential predictor roles in DLBCL, even though PKC-b expression is much lower in DLBCL cancers than either CLL or FL. Maybe this paragraph should be moved to the end of the section, to maintain the focus of the paper on PKC-b expression in DLBCL cancers.
We agree with the reviewer regarding further studies of PKC-β expression in CLL and FL. However, in the dataset we analyzed, survival data was not available for CLL and FL for more detailed analysis of PKC-β in these two subtypes, as indicated in the 2nd paragraph of the discussion section.
2. The authors claim that, "A similar association was evident with the IPI high-risk patients, but there was no discernable trend for the IPI low-risk group." This seems like a strong statement to make. From 5B, it seems that survival rates are the same for all expression levels in the IPI low-risk group, not that there is no discernable trend. Also, it is difficult to assert that the trend is the same in both the IPI intermediate and IPI high risk group, especially with the small sample sizes and insignificant p-value.
We agree with the reviewer and modified the wording in the sentence to indicate that the association between PKC-β expression and survival was not statistically significant in the high IPI group and there is completely lack of association in the low IPI group.
3. The authors claim that, "These results confirm that PKC-b is clinically important to BCR signaling." This is not necessarily true, unless one can show that it is the upregulation of PKC-b that is causing the other factors to be up-regulated, and not that PKC-b is just another byproduct of some other up-regulation.
Our statement is based on previous studies, for example, please refer to reference 9 for the role of PKC-β in BCR signaling.
4. To exclude the possibility of cross-hybridizations amplifying PKC-b expression signals the authors could check for the uniqueness of the probes to PKC-b on all of the arrays (both cDNA and affymetrix).
We agree with the reviewer. The issue has been addressed by detailed discussion of probe sequences in Table 2and the method section.
1. We suggest that the sentence be changed to reflect the gene expression from the microarrays – "Our analysis showed that the level of PKC-b expression was..."
2. We suggest that the sentence specify the expression of PKC-b – "...lymphoma (DLBCL), the level of PKC-b expression was significantly..."
3. Given that the p-value is significant, it doesn't seem necessary to also include it in a parenthetical reference.
1. To clarify the two PKC-b isoforms, it would be useful to change the format to something similar to – "The PKC-b1 and -b2 isoforms..."
2. There is no citation for "Additionally VEGF-mediated mitogenic effects were blocked by PKC-b inhibitor LY333531, but not by antisense oligonucleotides to PKC-a."
1. To clarify that the subtypes are being listed, we suggest using a colon after, "... among the three gene expression DLBCL subtypes: germinal ..."
2. On figure 3A, it is difficult to see some of the lines, so we suggest to bolden the Q3 lines.
3. On figure 3B, we suggest to make the PKC-b medium line bolder.
4. On figure 3B, we suggest to change the x-axis to years instead of months, to be consistent with the other graphs. Also, for consistency, it might be nice to see it divided into quartiles, instead of thirds.
5. On figure 5B, in all three plots, it is difficult to distinguish the Q3+Q4 line in the on-line version and would be much easier to read if this line were darker
1. In the first sentence"know" to "understand."
2. There is no citation for, "Global gene expression analysis with DNA microarray technology has proven to be a useful tool for target identification and validation in both the preclinical and clinical settings."
3. There is no citation for, "Gene expression analysis is currently used for diagnosis, prognosis and therapeutic decision-making in a variety of cancers."
4. Change "PKC-B" to "PKC-b"
5. Change from "drugable genes" to "genes that are potential drug targets."
6. Change to a colon after, "...clustered DLBCL patients into three subtypes: GCB, ABC..."
7. There is no reference to figure 2 when it is mentioned in, "Our analysis showed significantly higher PKC-b in the ABC..."
8. The reference should be to Table 3, not Table 2, after, "Gene annotation and pathway analyses presented in the current report..."
We have modified the manuscript to address every point in reviewer's comments, except for the suggestion to change the boldness of lines in Figure 3Band 5B. Since the figures were generated using the software MedCalc (see the Method's section), we could not modified each lines. However, since the lines are in color, we feel the readers should be able to distinguish different survival curves that were represented by different colors.
Reviewer's report 3
Dr. Matthew Coleman, Lawrence Livermore National Labs, Livermore, CA 94550. Nominated by Dr. Sandrine Dudoit.
Reject until major revisions are made. Scientific quality of the work is high and should be of general interest once major revisions are made. The paper would have been strengthened by multiple statistical testing of the hypotheses (see comments below). The tables and figures should be restricted to those that are critical. All other figures/tables should be moved to a supplemental data section.
General Comments for the Author:
The authors take a statistical and bioinformatics approach to rexamine the role for protein kinase C-beta (PKC-b) in tumor pathogenesis related to B-cell malignancies. They reanalyzed global gene expression data to correlate gene expression across diagnostic groups and than determined associations between PKC-b expression and survival. The paper is clearly written and the scientific data is very convincing regarding PKC-b expression. However the discussion/scientific impact only corroborate previous expression studies on the role of multiple critical cellular pathways such as anti-apoptosis, cell proliferation, and signalling pathways, that were found to be loosely associated with PKC-b function. A more rigorous bioinformatics study on the relationship between PKC-b expression and the other pathways would have added impact to the study. Multiple tools are available on the web for looking at co-expression and pathway analysis and should be used to strength the authors findings. Many of these tools would also add significance the pathway/association analysis, which was found missing from this study. The elevated PKC-b levels tied to prognosis is an interesting finding that should be of interest to the general readership.
We appreciate that the reviewer pointed out "Scientific quality of the work is high and should be of general interest" and "The paper is clearly written and the scientific data is very convincing". We have made revisions to address reviewer's comments. Similar to our response to the 2nd reviewer's report, we modified the abstract and the discussion section to highlight the novelty of the study. We have performed additional analysis to identify genes with expressions highly correlated with PKC-β, followed by mapping them to biological pathways. These analyses confirmed our results in Figure 6 and Table 3. Please see below for detailed response to each of reviewer's specific comments.
1. Genes not represented between two different population and different array platforms or different analysis algorithms should at have identify similar pathways if not exact genes. (17 vs. 13 genes). Is this possible with your findings? Also, the authors could reanalyze this data starting from raw microarray data to see if similar algorithms derive the same answer.
The issue has been widely studied recently and it has been demonstrated that even though different gene sets were used for prognostication in cancer patients, they showed significant agreement in the outcome predictions and are probably tracking a common set of biologic phenotypes. Please see a recent publication on New England Journal of Medicine: Fan C, Oh DS, Wessels L, Weigelt B, Nuyten DS, Nobel AB, van't Veer LJ, Perou CM. (2006) Concordance among gene-expression-based predictors for breast cancer. N Engl J Med. 355(6):560-9. This topic is beyond the scope of the current study.
2. How was the data sorted into quartiles? It appears arbitrary and not statistical as written. Would the use of supervised and unsupervised methods distinguish between the same groupings of data and would these methods validate the significance?
In our study, we applied two approaches in analyzing the correlation between PKC-β expression and patient survival. In one approach, we used the Kaplan-Meier survival analysis after grouping patients into 4 quartiles. We also applied Cox Proportional-Hazards Regression model to analyze the same dataset and reached the same conclusion. However, we only reported the Kaplan-Meier analysis results in the original manuscript. To address reviewer's concern, we added a sentence in the result section to indicate that the Kaplan-Meier analysis results were confirmed by the Cox Proportional-Hazards Regression model.
3. Table two should be moved to supplemental data
We also considered the option of moving Table 2 to a supplemental file. However, in the 2nd reviewer's report, the reviewer pointed out the importance of analyzing probe sequences in both cDNA and Affymetrix array platforms to exclude the possibility of cross hybridization. Therefore, we feel it is necessary to show our analysis and discussion of probe sequences in Table 2.
4. Table 3 gene list should be summarized to include significance based on comparing to background array data. The individual gene list could be moved to supplemental data.
Table 3 has been modified to only include gene categories. The individual gene list has been moved into Additional file 1. The main text has also been modified accordingly.
5. Figures 2, and 5A are not necessary.
Figure 2 and 5A provide a direct visualization of the distribution of different PKC-β quartiles in DLBCL subtypes. We feel that it is better to describe the results using both figures and text.
6. Figure 4 is misaligned obscuring the y axis.
We appreciate the reviewer pointed this out and it has been fixed.
7. Graphs should be redone without color to better illustrate the data.
As we discussed in response to the 2nd reviewer's report, the figures were generated using the software MedCalc (see the Method's section). We feel the readers should be able to better distinguish different survival curves that were represented by different colors.