Transcriptomic Analysis of the Effect of Metformin against Cisplatin-Induced Ototoxicity: A Potential Mechanism of Metformin-Mediated Inhibition of Thioredoxin-Interacting Protein (Txnip) Gene Expression
Round 1
Reviewer 1 Report
Though widely used in chemotherapies, cisplatin is ototoxic and may cause irreversible sensorineural loss in patients. Understanding the molecular mechanisms of cisplatin ototoxicity and finding new ways to prevent or alleviate this adverse side effect will greatly improve the life quality of patients. In this manuscript “Transcriptomic analysis of the effect of metformin against cisplatin induced ototoxicity: a potential mechanism of metformin-mediated inhibition of Thioredoxin-interacting protein (Txnip) gene expression”, the authors presented the transcriptomic analysis of HEI-OC1 cells after cisplatin treatment with/without metformin pre-conditioning, analyzed the pathways affected under each conditions and speculated that metformin protected HEI-OC1 cells against cisplatin ototoxicity through inhibition of Txnip expression. This work provides new insights into cisplatin ototoxicity and its protection; however, the impact of this work is compromised by a few major concerns.
First of all, HEI-OC1 cells was used as the sole model for transcriptomic analysis and pathway enrichment analysis. HEI-OC1 cell line is an immortalized cell line derived from mouse cochlea and is the most used cell line for ototoxicity research. But HEI-OC1 cells are definitely different from the primary cochlear cells in morphology, differentiation state and gene transcriptions. Thus, discoveries made with HEI-OC1 cells should be validated with primary cochlear cells. Such validation is missing in current manuscript, raising concerns whether the results from this study reflect the changes in primary tissues.
Second, the usage of Microarray is also a concern. Microarray probes are designed based on prior knowledge of the transcriptome in cells, and inevitably such a prior knowledge-based technique introduces bias. In contrast, RNAseq is a better technique without the requirement of prior knowledge. Actually, multiple RNAseq datasets collected from cisplatin-treated HEI-OC1 cells have been published. A comparison of the microarray data from this study and the published RNAseq data should be performed so audience will have a better idea how reliable the current study is.
Third, the analyses of the transcriptomic changes are superficial. The genes with significant expression difference in each comparison were listed, and enrichment of pathways and GO terms was performed. Information from such analyses is very limited, and the data should be analyzed at higher dimensions. For example, how the expression of a particular gene changes at different time points? Whether there are some genes transiently up- or down-regulated by cisplatin or by cisplatin+metformin? How many up-regulated genes by cisplatin is down-regulated by metformin or vice versa? Figure 4 in the manuscript shows how metformin induced genes were changed by cisplatin, but it is of interest for readers to know how cisplatin-induced genes were modulated under the presence of metformin. In addition, it would be better to show a combined heatmap of all conditions at all time points after unsupervised clustering to show gene modules respond differently to treatments at different timepoints.
Besides, the authors proposed a mechanism focusing on the expression of Txnip. This mechanism is based on Txnip expression changes in this study and in previous published results from other groups. This mechanism might be a potential target for cisplatin ototoxicity protection, but it requires further experimental validation, especially when this finding is contradicting to observations in literatures. Thus, I personally think it should not be emphasized in the title to avoid potential misinterpretation.
In addition to the aforementioned major concerns, there are other issues with the manuscript:
1) In figure 1Bd, statistic comparisons should be focused on cisplatin vs cisplatin+metformin instead of between different caspases.
2) For differential expressed genes, adjusted p value instead of p value should be used as multiple comparisons were conducted.
3) In figure 2B, heatmaps should be replotted. Genes names on the right are not readable, so remove them. Colors are saturated so no information can be extracted other than up- or down-regulated.
4) For figure 2C, change to volcano plots to show fold change and p value simultaneously.
5) Fig 3, adjust p value should be used.
6) Raw data and processed data for all the genes detected by microarray should be provided as supplementary tables.
Author Response
Reviewer 1 Though widely used in chemotherapies, cisplatin is ototoxic and may cause irreversible sensorineural loss in patients. Understanding the molecular mechanisms of cisplatin ototoxicity and finding new ways to prevent or alleviate this adverse side effect will greatly improve the life quality of patients. In this manuscript “Transcriptomic analysis of the effect of metformin against cisplatin induced ototoxicity: a potential mechanism of metformin-mediated inhibition of Thioredoxin-interacting protein (Txnip) gene expression”, the authors presented the transcriptomic analysis of HEI-OC1 cells after cisplatin treatment with/without metformin pre-conditioning, analyzed the pathways affected under each conditions and speculated that metformin protected HEI-OC1 cells against cisplatin ototoxicity through inhibition of Txnip expression. This work provides new insights into cisplatin ototoxicity and its protection; however, the impact of this work is compromised by a few major concerns. First of all, HEI-OC1 cells was used as the sole model for transcriptomic analysis and pathway enrichment analysis. HEI-OC1 cell line is an immortalized cell line derived from mouse cochlea and is the most used cell line for ototoxicity research. But HEI-OC1 cells are definitely different from the primary cochlear cells in morphology, differentiation state and gene transcriptions. Thus, discoveries made with HEI-OC1 cells should be validated with primary cochlear cells. Such validation is missing in current manuscript, raising concerns whether the results from this study reflect the changes in primary tissues.
--> Thank you for your comment. I get your point and agree to your concern to some point. Our experiment was based on the HEI-OC1 cells only for the transcriptomic analysis; and the changes in primary tissues might be different. However, transcriptomic changes in the auditory cell line would give certain information regarding the alteration of genes after the cisplatin exposure which would be used as reference data, and further, out study would provide details regarding time-dependent gene alternation. Nevertheless, I mentioned the limitation of our study in the discussion as you’ve mentioned. Thank you.
Line 22 on page 13: Although, our experiment was performed on HEI-OC1 cell line and further validation studies are required in the future,
Second, the usage of Microarray is also a concern. Microarray probes are designed based on prior knowledge of the transcriptome in cells, and inevitably such a prior knowledge-based technique introduces bias. In contrast, RNAseq is a better technique without the requirement of prior knowledge. Actually, multiple RNAseq datasets collected from cisplatin-treated HEI-OC1 cells have been published. A comparison of the microarray data from this study and the published RNAseq data should be performed so audience will have a better idea how reliable the current study is.
--> Thank you for your comment. We are aware of the difference between RNA-seq and microarray, and we prefer RNA-seq these days in our lab. But this experiment was performed with microarray in several different time points (6, 15, 24, and 48 hr). I’ve searched for several published papers reporting the cisplatin ototoxicity with RNA-seq. Results were similar and not much different from our study, and I’ve updated the finding in the discussion section as below.
There are several reports demonstrating the gene expression change in cisplatin-induced ototoxicity in HEI-OC1 cells using RNA sequencing analysis [24, 25]. In a study evaluating the gene expression change after 30 hr cisplatin exposure, downregulated DEGs were associated with several KEGG pathways such as ‘systemic lupus erythematosus’, ‘alcoholism’ ‘viral carcinogenesis’, ‘PI3K-Akt signaling pathway’, ‘Rap1 signaling pathway’, and ‘HIF-1 signaling pathway’; which are in accord with our data (Supplementary table 2). They also reported that upregulated DEGs were significantly associated with autophagy, apoptosis-associated processes, response to DNA damage and cell cycle arrest; and demonstrated several genes such as phorbol-12-myristate-13-acetate-induced protein 1 (PMAIPI), Bcl 2 binding component 3 (BBC3), zinc finger matrin type 3 (ZMAT3), p53 induced death domain protein 1 (PIDD1), B-cell translocation gene protein 2 (BTG2), thioredoxin-inter-acting protein (TXNIP), DNA damage induced apoptosis suppressor (DDIAS), cycle-dependent kinase inhibitor 1a (Cdkn1a), transformation related protein 53-induced nuclear protein 1 (Trp53inp1), Foxo3, and Fas; which are also similar to our results.
24. Gaun,G; He, X; Chen, J; Bin, L; Tang, X. Identifying the mechanisms underlying the protective effect of tetramethylpyrazine against cisplatin induced in vitro ototoxicity in HEI OC1 auditory cells using gene expression profiling. Molecular medicine reports. 2020. 22: 5033-5068. doi: 10.3892/mmr.2020.11631.
25. Liu, C; Zheng; Z; Li, W; Tang, D; Zhao, L; He, Y; Li, H. Inhibition of KDM5A attenuates cisplatin-induced hearing loss via regulation of the MAPK/AKT pathway. Cell Mol Life Sci. 2022. 79(12):596. doi: 10.1007/s00018-022-04565-y.
Third, the analyses of the transcriptomic changes are superficial. The genes with significant expression difference in each comparison were listed, and enrichment of pathways and GO terms was performed. Information from such analyses is very limited, and the data should be analyzed at higher dimensions. For example, how the expression of a particular gene changes at different time points?
-->We did not include the 6 hr and 15 hr time points for further analysis because the number of total DEGs in 6hr and 15 hr was not sufficient for further analysis. So, we’ve focused on later time points in our experiment. We’ve selected several key genes and identified their time dependent expression by RT-PCR (Fig 5).
Whether there are some genes transiently up- or down-regulated by cisplatin or by cisplatin+metformin?
-->I’m not sure I’ve understood your point but if you are mentioning the genes that were down-regulated or up-regulated at 24 hr time point and vice versa at 48 hr time points; yes, there are some genes that moved oppositely. When I evaluated and compared the genes in 24 hr and 48 hr of metformin pretreated cisplatin group, there are 2 or 3 genes that are statistically significant. However, in our study we evaluated and analyzed genes that moved oppositely in response to cisplatin, because we want to evaluated the effect of metformin in cisplatin induced ototoxicity. That is, we compared genes that moved oppositely in cisplatin group and metformin-pretreated group at same time points. These results are shown in figure 4, table 7 and supplementary table 3.
How many up-regulated genes by cisplatin is down-regulated by metformin or vice versa? Figure 4 in the manuscript shows how metformin induced genes were changed by cisplatin, but it is of interest for readers to know how cisplatin-induced genes were modulated under the presence of metformin.
--> Indeed, Fig 4 explains which genes were up or downregulate by metformin-pretreatment. If you look at Fig 4A, the right colume is comparing genes between 24 metformin-pretreated and then cisplatin applied for 24 hr group vs cisplatin applied for 24hr group; which means metformin induced up-regulation of these genes in cisplatin applied condition. The left colume in Fig 4A is comparing genes between in cisplatin applied for 24 hr group vs control. We did demonstrate how cisplatin-induced genes were modulated under the presence of metformin which would be of interest for readers.
In addition, it would be better to show a combined heatmap of all conditions at all time points after unsupervised clustering to show gene modules respond differently to treatments at different timepoints.
--> We could not include data of 6 hr and 15 hr time points for further analysis because the number of total DEGs in 6hr and 15 hr was not sufficient for further analysis.
Besides, the authors proposed a mechanism focusing on the expression of Txnip. This mechanism is based on Txnip expression changes in this study and in previous published results from other groups. This mechanism might be a potential target for cisplatin ototoxicity protection, but it requires further experimental validation, especially when this finding is contradicting to observations in literatures. Thus, I personally think it should not be emphasized in the title to avoid potential misinterpretation.
-> I do get point. But I’d like to keep the title, since we’ve suggested Txnip as a “potential mechanism” in the title.
In addition to the aforementioned major concerns, there are other issues with the manuscript: 1) In figure 1Bd, statistic comparisons should be focused on cisplatin vs cisplatin+metformin instead of between different caspases. We’ve compared the data and added the statistics in the figure 1Bd. 2) For differential expressed genes, adjusted p value instead of p value should be used as multiple comparisons were conducted. 5) Fig 3, adjust p value should be used. Actually, all the data we’ve chose were based on false discovery rate. We’ve changed the legend to FDR < 0.01 in Fig 3, and table 3-7. 3) In figure 2B, heatmaps should be replotted. Genes names on the right are not readable, so remove them. Colors are saturated so no information can be extracted other than up- or down-regulated. I’ve erased gene names on the right since they are not readable. However, we regret to say, but we cannot change the gradation color of the figure 2B at this time. 4) For figure 2C, change to volcano plots to show fold change and p value simultaneously. I apologize but we cannot change the plot to volcano plot at this time. 6) Raw data and processed data for all the genes detected by microarray should be provided as supplementary tables. I’ve added supplementary tables. Thank you.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors studied an inner ear cell line sensitive to cisplatin and how metformin could prevent this ototoxicity, and its dependency on time of metformin exposure. The results show increased gene expression alteration with increased time and that this effect could be partwise attenuated by metformin. In conclusion, gene expression change due to cisplatin may be attenuated by metformin.
TRhe study os logical and I see no flaws. Or even typos!
Author Response
Thank you for your review.
Reviewer 3 Report
The manuscript presents findings of research on the effect of metformin against cisplatin induced ototoxicity on work studied by the standard methods. The paper presents very interesting results as well as an inquisitive and reliable interpretation of the research results.
Author Response
Thank you for your review.
Round 2
Reviewer 1 Report
The majority of my concerns were solved in the revised manuscript; however, there are still problems with the analyses and presentation of the data. Particularly for Figure 2 and Figure 4.
Figure 2B, all heatmaps are saturated because data from a single comparison were normalized and plotted. Essential information for the gene expression along the timeline was not conveyed through such presentations. A better way to analyze the data is: 1) get a list of differentially expressed genes for all the time points and all comparisons (including DEGs at 6 hours and 15 hours for both cisplatin vs control and metformin+cisplatin vs cisplatin), and then 2) plot the heatmap for all the timepoints and conditions after normalization and scaling. As references, check out the Figure 4A from Kang et al., 2019 (https://www.frontiersin.org/articles/10.3389/fcell.2019.00204/full), or Figure 1I from Diaz et al., 2020 (https://elifesciences.org/articles/52707). Analyzing the data with all timepoints and condition together will reveal more information for the readers, such as modules of genes behaving similarly across the timeline, transiently up- or down-regulated genes, and time-dependent up- or down-regulation.
Figure 4, the authors showed how metformin+cisplatin induced DEGs were changed by cisplatin alone. For the purpose of demonstration how metformin provides protection, it’s better to show how cisplatin-induced DEGs respond to metformtin+cisplatin. In another word, instead of grouping genes on metformin+cisplatin vs cisplatin, genes should be grouped based on cisplatin vs control, and then show their changes in metformin+cisplatin vs cisplatin.
In addition to those remaining problems, there are two more issues. The numbers of biological replicates for qPCR and microarray were not mentioned in either method section or result text. Microarray data for metformin alone sample is not presented, which should be discussed as a caveat for the experimental design.
Author Response
Figure 2B, all heatmaps are saturated because data from a single comparison were normalized and plotted. Essential information for the gene expression along the timeline was not conveyed through such presentations. A better way to analyze the data is: 1) get a list of differentially expressed genes for all the time points and all comparisons (including DEGs at 6 hours and 15 hours for both cisplatin vs control and metformin+cisplatin vs cisplatin), and then 2) plot the heatmap for all the timepoints and conditions after normalization and scaling. As references, check out the Figure 4A from Kang et al., 2019 (https://www.frontiersin.org/articles/10.3389/fcell.2019.00204/full), or Figure 1I from Diaz et al., 2020 (https://elifesciences.org/articles/52707). Analyzing the data with all timepoints and condition together will reveal more information for the readers, such as modules of genes behaving similarly across the timeline, transiently up- or down-regulated genes, and time-dependent up- or down-regulation.
<Response> Thank you for your valuable advice. I’ve checked references you’ve mentioned, and I find them very interesting. But I have two things to point out. First, we have focused on 24 hr and 48 hr because (1) all caspases seemed to have meaningful differences at 24 hr and 48 hr conditions; (2) the number of total DEGs in 6 hr and 15 hr was not sufficient for further analysis. Second, if we analyze a list of DEG from all the time points at this moment, I’m afraid that our other data and following results in manuscript have to be re-written all over. I think and suggest that our data has a strength in that it analyzed the gene expression in 24 hr and 48 hr. Maybe I can change Fig 2B as below instead of previous heatmaps.
Figure 4, the authors showed how metformin+cisplatin induced DEGs were changed by cisplatin alone. For the purpose of demonstration how metformin provides protection, it’s better to show how cisplatin-induced DEGs respond to metformtin+cisplatin. In another word, instead of grouping genes on metformin+cisplatin vs cisplatin, genes should be grouped based on cisplatin vs control, and then show their changes in metformin+cisplatin vs cisplatin.
<Response> I get your point. It could have been better to groups genes based on cisplatin vs control, and then show their changes in metformin+cisplatin vs cisplatin as you’ve mentioned. But, likewise, I apologize but cannot re-analyze the data since it would change the result of network analysis and following qPCR.
In addition to those remaining problems, there are two more issues. The numbers of biological replicates for qPCR and microarray were not mentioned in either method section or result text. Microarray data for metformin alone sample is not presented, which should be discussed as a caveat for the experimental design.
<Response> The number of biological replicated for qPCR are written in method section (2.7 Quantitative real-time PCR, line 6) and in figure 5 legend (N=5). Also, I’ve changed Fig 2B to show the expression of metformin at 24 hr and 48 hr.
Author Response File: Author Response.docx
