Integrative transcriptomic analysis of tissue-specific metabolic crosstalk after myocardial infarction
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This paper will be of considerable interest to researchers studying the interactions between metabolic responses in myocardial infarction. Ultimately increased understanding of these interactive metabolic responses could lead to exploration of new avenues of treatment.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Myocardial infarction (MI) promotes a range of systemic effects, many of which are unknown. Here, we investigated the alterations associated with MI progression in heart and other metabolically active tissues (liver, skeletal muscle, and adipose) in a mouse model of MI (induced by ligating the left ascending coronary artery) and sham-operated mice. We performed a genome-wide transcriptomic analysis on tissue samples obtained 6- and 24 hr post MI or sham operation. By generating tissue-specific biological networks, we observed: (1) dysregulation in multiple biological processes (including immune system, mitochondrial dysfunction, fatty-acid beta-oxidation, and RNA and protein processing) across multiple tissues post MI and (2) tissue-specific dysregulation in biological processes in liver and heart post MI. Finally, we validated our findings in two independent MI cohorts. Overall, our integrative analysis highlighted both common and specific biological responses to MI across a range of metabolically active tissues.
Article activity feed
-
-
Author Response:
Reviewer #1:
The authors note how previous studies on myocardial infarction have usually studied individual tissues and not examined the cross talk between tissues and their dysregulation. To address this challenge they have therefore performed, in a mouse model of MI, an integrated analysis of heart, liver, skeletal muscle and adipose tissue responses at 6 and 24 hours. They have then validated their findings at 24 hours in two independent mouse model data sets.
A major strength is their comprehensive approach. They have used high throughput RNA seq and applied integrative network analysis. They show for multiple genes whether they are up regulated or down regulated in these four tissues at the 6 and 24 hour time points and whether the regulation directions are concordant or opposite and note in particular that …
Author Response:
Reviewer #1:
The authors note how previous studies on myocardial infarction have usually studied individual tissues and not examined the cross talk between tissues and their dysregulation. To address this challenge they have therefore performed, in a mouse model of MI, an integrated analysis of heart, liver, skeletal muscle and adipose tissue responses at 6 and 24 hours. They have then validated their findings at 24 hours in two independent mouse model data sets.
A major strength is their comprehensive approach. They have used high throughput RNA seq and applied integrative network analysis. They show for multiple genes whether they are up regulated or down regulated in these four tissues at the 6 and 24 hour time points and whether the regulation directions are concordant or opposite and note in particular that for the liver both concordant and opposite effects occur. They identify key tissue specific clusters in each tissue and identify the key genes in each cluster. Finally they use whole body modelling to identify cross talk between tissues.
A further strength of this paper is the integration of transcriptomic data (differential expression, functional analysis and reporter metabolite analysis). The final strength is the very clear presentation of the findings and their implications such that the reader gets a very clear message and at the same time can go in to more detail if this is their area of research interest.
There are no major weaknesses. The authors have achieved their aims and the data supports their conclusions.
This work represents a major advance in both methodology and understanding of a multi tissues approach to the study of the metabolic impact of MI and the underlying up and down regulation of relevant genes.
The relevance of these findings in human MI will need to be tested and may ultimately have therapeutic implications.
First and foremost, we would like to thank the reviewer for the positive and encouraging comments. We agree that further research, especially rigorous validation of the findings from this work in humans, is needed and hopefully it can be translated into clinical settings. Moreover, we would like to thank the reviewer for his highlight on our comprehensive approach that we hope can be a framework for future multi-tissue research in disease setting.
Reviewer #2:
The authors collected post-myocardial infarction (MI) transcriptome data from a mouse model as well as sham-operated control mice to identify systemic molecular changes in multiple tissues at pathway level. The data were collected at two time points (6 hours and 24 hours post-MI), and several computational systems biology tools were applied to the dataset to identify altered molecular processes. The applied tools vary from very standard tools (eg. enrichment analysis) to advanced methods based on mapping data on biological networks. A specific focus was put on the altered signaling pathways as well as metabolic pathways and metabolites. Identified up-/down-regulated pathways were in agreement with the literature.
Strengths:
- One unique aspect of the work is the fact that the transcriptomic data were collected from not only heart, the source tissue for MI, but also from three more tissues (liver, skeletal mouse, adipose). Therefore, molecular alterations in the related tissues were also able to be monitored and discussed comparatively. The introduced transcriptomic dataset has a high re-use potential by other researchers in the field since coverage of responses by four tissues at two different time points makes it unique.
- Correlation-based coexpression networks were created for all four tissues, and some of the clusters in these networks were shown to be tissue-specific clusters, which nicely validates both the experimental and computational approach in the paper.
- The results were validated by using independent transcriptomic datasets available in the literature. The authors showed that there is a high overlap between their dataset and the literature datasets in terms of identified differentially expressed genes and enriched pathways. This additional validation strengthens the results reported in the manuscript.
- Use of a variety of computational approaches and showing that they point to similar or complementary molecular mechanisms increase the impact of the paper. The employed computational tools include not only information-extraction methods such as enrichment, coexpression networks, reporter metabolites, but also predictive methods based on modelling. The authors construct a multi-tissue genome-scale metabolic network covering all four tissues of interest in the study, and they show that this model can correctly predict some major post-MI changes in the metabolism. It is interesting to see that two completely different computational approaches (constraint-based metabolic modeling versus information-extraction based approaches) point to same/similar molecular mechanisms.
We would like to thank the reviewer for providing positive comments and a comprehensive summary of our work. We also really appreciate the constructive comments from the reviewer to improve our work.
Weaknesses:
- Regarding predictions made by multi-tissue metabolic network modeling, the control case fluxes were predicted by maximizing the rate of lipid droplet accumulation in the adipose tissue. Although there is an agreement between the model predictions and the results obtained by other bioinformatics tools used in the study as well as literature information, it looks rather oversimplification to assume that all other three tissues are programmed to serve for maximum fat production in adipose tissue. This should be further elaborated by the authors.
We would like to thank the reviewer for the comment and we agree that there is a simplification of the situation in the modeling. However, we would like also to emphasize that the model has been carefully constrained with the dietary composition as well as the tissue specific resting energy expenditure. In our opinion, these constraints have already included a great part of the metabolic activity and satisfied the basic metabolic needs of the mice. The rest of the energy in the diet could be either used as physical activity (energy production in muscle) or stored as fat (lipid droplet accumulation in adipose tissue), and in our analysis, we assumed the latter as we think it is more realistic in this study as mice in the cage might have very little physical activities. We added a clarification for this in the revised manuscript as follows,
‘To simulate the metabolic flux distribution in the sham-operated mice, we set the lipid droplet accumulation reaction in adipose tissue (m3_Adipose_LD_pool) as the objective function as we assume the energy additional to the resting energy expenditure will be mostly stored as fat rather than used by the muscle for physical activities because mice raised in the cages might have very little exercise. Then, we used parsimonious FBA to calculate the flux distribution.’
Reviewer #3:
In the manuscript, "Integrative transcriptomic analysis of tissue-specific metabolic crosstalk after myrocardial infarction" by Arif et al., the authors describe analyses of transcriptomes of +/- myocardial infarction (MI) mice. The study is useful and reports interesting results. These results could be of interest to further develop cellular insight in effects and treatments for MI. However, I do not find any methodological advances here. The manuscript appears to be a repository of transcriptomics analyses. All the techniques used have been tried and applied to other scientific problems. The authors have presented differential expression analysis, followed by GSEA, and then they perform different network analyses - co-expression networks, reporter analyses, multi-tissue model, etc.
My main issues are that the authors do too many different analyses but neither of them get sufficient light in the paper. Also no other independent quantitative evidence is shown in support of results of their analyses. Further, validation was done the same way the pipeline was built. This makes their results comes across as circular. For e.g. when validating metabolic models of cells built using transcriptomic data, CRISPR-Cas9 essentiality screens are used. Here, they basically repeated the same analyses on the same transcriptome from a different experiment it appears.
First of all, we would like to thank the reviewer for the positive summary of our research. We agree that this study can be useful to be explored further, especially by validating it in human. We also would like to thank you for the constructive comments. We agree that we presented multiple transcriptomics analyses that have been used before. Apart from understanding the metabolic effect of MI in multiple tissues (which is unique as of now), our secondary goal is to propose a novel integrative framework for analyzing multi-tissue transcriptomics data based on the available techniques. We would like to emphasize that, even though the singular analyses were not novel, the integrative analysis in multi-tissue and disease setting both at transcriptomic and metabolic crosstalk level is a strong novelty of this study. This required employing not only state-of-art network analyses but also reconstruction of multi-tissue models through new methods that enable joint modeling of the metabolic interactions within and between tissues.
As this study is unique (as of now), we tried our best to validate it with other data with similar settings (from a tissue and we found only transcriptomics data) and run our pipeline to validate and strengthen our findings. Moreover, we also recognized the limitation that all the results presented in this study are purely based on transcriptomics data (as stated in the “Discussion” section of the manuscript). More experiments, such as with metabolomics and proteomics data, are in our pipeline to complement the results from the current study. In summary, we recognized the reviewer’s concerns and we would like to address it in our future studies.
-
Reviewer #3 (Public Review):
In the manuscript, "Integrative transcriptomic analysis of tissue-specific metabolic crosstalk after myrocardial infarction" by Arif et al., the authors describe analyses of transcriptomes of +/- myocardial infarction (MI) mice. The study is useful and reports interesting results. These results could be of interest to further develop cellular insight in effects and treatments for MI. However, I do not find any methodological advances here. The manuscript appears to be a repository of transcriptomics analyses. All the techniques used have been tried and applied to other scientific problems. The authors have presented differential expression analysis, followed by GSEA, and then they perform different network analyses - co-expression networks, reporter analyses, multi-tissue model, etc.
My main issues are that …
Reviewer #3 (Public Review):
In the manuscript, "Integrative transcriptomic analysis of tissue-specific metabolic crosstalk after myrocardial infarction" by Arif et al., the authors describe analyses of transcriptomes of +/- myocardial infarction (MI) mice. The study is useful and reports interesting results. These results could be of interest to further develop cellular insight in effects and treatments for MI. However, I do not find any methodological advances here. The manuscript appears to be a repository of transcriptomics analyses. All the techniques used have been tried and applied to other scientific problems. The authors have presented differential expression analysis, followed by GSEA, and then they perform different network analyses - co-expression networks, reporter analyses, multi-tissue model, etc.
My main issues are that the authors do too many different analyses but neither of them get sufficient light in the paper. Also no other independent quantitative evidence is shown in support of results of their analyses. Further, validation was done the same way the pipeline was built. This makes their results comes across as circular. For e.g. when validating metabolic models of cells built using transcriptomic data, CRISPR-Cas9 essentiality screens are used. Here, they basically repeated the same analyses on the same transcriptome from a different experiment it appears.
-
Reviewer #2 (Public Review):
The authors collected post-myocardial infarction (MI) transcriptome data from a mouse model as well as sham-operated control mice to identify systemic molecular changes in multiple tissues at pathway level. The data were collected at two time points (6 hours and 24 hours post-MI), and several computational systems biology tools were applied to the dataset to identify altered molecular processes. The applied tools vary from very standard tools (eg. enrichment analysis) to advanced methods based on mapping data on biological networks. A specific focus was put on the altered signaling pathways as well as metabolic pathways and metabolites. Identified up-/down-regulated pathways were in agreement with the literature.
Strengths:
One unique aspect of the work is the fact that the transcriptomic data were collected …
Reviewer #2 (Public Review):
The authors collected post-myocardial infarction (MI) transcriptome data from a mouse model as well as sham-operated control mice to identify systemic molecular changes in multiple tissues at pathway level. The data were collected at two time points (6 hours and 24 hours post-MI), and several computational systems biology tools were applied to the dataset to identify altered molecular processes. The applied tools vary from very standard tools (eg. enrichment analysis) to advanced methods based on mapping data on biological networks. A specific focus was put on the altered signaling pathways as well as metabolic pathways and metabolites. Identified up-/down-regulated pathways were in agreement with the literature.
Strengths:
One unique aspect of the work is the fact that the transcriptomic data were collected from not only heart, the source tissue for MI, but also from three more tissues (liver, skeletal mouse, adipose). Therefore, molecular alterations in the related tissues were also able to be monitored and discussed comparatively. The introduced transcriptomic dataset has a high re-use potential by other researchers in the field since coverage of responses by four tissues at two different time points makes it unique.
Correlation-based coexpression networks were created for all four tissues, and some of the clusters in these networks were shown to be tissue-specific clusters, which nicely validates both the experimental and computational approach in the paper.
The results were validated by using independent transcriptomic datasets available in the literature. The authors showed that there is a high overlap between their dataset and the literature datasets in terms of identified differentially expressed genes and enriched pathways. This additional validation strengthens the results reported in the manuscript.
Use of a variety of computational approaches and showing that they point to similar or complementary molecular mechanisms increase the impact of the paper. The employed computational tools include not only information-extraction methods such as enrichment, coexpression networks, reporter metabolites, but also predictive methods based on modelling. The authors construct a multi-tissue genome-scale metabolic network covering all four tissues of interest in the study, and they show that this model can correctly predict some major post-MI changes in the metabolism. It is interesting to see that two completely different computational approaches (constraint-based metabolic modeling versus information-extraction based approaches) point to same/similar molecular mechanisms.
Weaknesses:
- Regarding predictions made by multi-tissue metabolic network modeling, the control case fluxes were predicted by maximizing the rate of lipid droplet accumulation in the adipose tissue. Although there is an agreement between the model predictions and the results obtained by other bioinformatics tools used in the study as well as literature information, it looks rather oversimplification to assume that all other three tissues are programmed to serve for maximum fat production in adipose tissue. This should be further elaborated by the authors.
-
Reviewer #1 (Public Review):
The authors note how previous studies on myocardial infarction have usually studied individual tissues and not examined the cross talk between tissues and their dysregulation. To address this challenge they have therefore performed, in a mouse model of MI, an integrated analysis of heart, liver, skeletal muscle and adipose tissue responses at 6 and 24 hours. They have then validated their findings at 24 hours in two independent mouse model data sets.
A major strength is their comprehensive approach. They have used high throughput RNA seq and applied integrative network analysis. They show for multiple genes whether they are up regulated or down regulated in these four tissues at the 6 and 24 hour time points and whether the regulation directions are concordant or opposite and note in particular that for the …
Reviewer #1 (Public Review):
The authors note how previous studies on myocardial infarction have usually studied individual tissues and not examined the cross talk between tissues and their dysregulation. To address this challenge they have therefore performed, in a mouse model of MI, an integrated analysis of heart, liver, skeletal muscle and adipose tissue responses at 6 and 24 hours. They have then validated their findings at 24 hours in two independent mouse model data sets.
A major strength is their comprehensive approach. They have used high throughput RNA seq and applied integrative network analysis. They show for multiple genes whether they are up regulated or down regulated in these four tissues at the 6 and 24 hour time points and whether the regulation directions are concordant or opposite and note in particular that for the liver both concordant and opposite effects occur. They identify key tissue specific clusters in each tissue and identify the key genes in each cluster. Finally they use whole body modelling to identify cross talk between tissues.
A further strength of this paper is the integration of transcriptomic data (differential expression, functional analysis and reporter metabolite analysis). The final strength is the very clear presentation of the findings and their implications such that the reader gets a very clear message and at the same time can go in to more detail if this is their area of research interest.
There are no major weaknesses. The authors have achieved their aims and the data supports their conclusions.
This work represents a major advance in both methodology and understanding of a multi tissues approach to the study of the metabolic impact of MI and the underlying up and down regulation of relevant genes.
The relevance of these findings in human MI will need to be tested and may ultimately have therapeutic implications.
-
Evaluation Summary:
This paper will be of considerable interest to researchers studying the interactions between metabolic responses in myocardial infarction. Ultimately increased understanding of these interactive metabolic responses could lead to exploration of new avenues of treatment.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
-
-
-