Decoding epidermis reconstruction through time-course multi-omics data
Podium 45
Presented by: Elias Bou Samra
The development of epidermis and its homeostasis require tightly regulated epidermal keratinocyte proliferation, differentiation and apoptosis, and these processes are governed by cooperation and interaction of molecules, namely transcripts, proteins, and metabolites. Each of these molecules provides a unique, complementary, and partly independent view of the genome and hence embed essential information about the regulatory mechanisms orchestrating epidermis development. The complexity of these regulations makes it clear that probing the system at a single molecular level gives only limited information, hence integrating and inferring regulatory interactions from these data is crucial.
Methods: To investigate the sequential events leading to epidermis differentiation at a multi-molecular level, we performed concurrent longitudinal high-throughput RNA-seq, miRNA-Seq, quantitative proteomics and metabolomics in a commercial reconstructed human epidermis model (SkinEthic Episkin Co, Lyon, France). In this regard, datasets were collected at different timepoints over twenty-eight days of culture. To identify complex relationships between omics layers, an analytic and integration framework was used for multi-omics longitudinal datasets that relies on multi-omics kinetic clustering using the timeOmics approach [1] and multi-layer network-based analysis using both data-driven and knowledge-driven building methods [2].
Results: Following data pre-processing, the molecules that varied most significantly during epidermis reconstruction were selected. Those with similar expression patterns over time were first modelled and clustered together. They showed distinctive temporal patterns characterized by their initial levels of abundance and variation during differentiation. For example, for molecules expressed on both gene and protein layers, temporal delay between mRNA formation and protein synthesis could be detected. In addition, enrichment analyses of GO terms performed on genes and proteins of each cluster recapitulated relevant biological functions previously described in skin reconstruction studies [3-5].
Next, by keeping the clustering information, a 4 layer network was built: two gene layers build from mRNA co-expression and miRNA-mRNA predictive regulations, a protein-protein interaction layer build from measured proteins and BioGRID known interactions, and a metabolite layer from KEGG pathways. Not only does this multi-layers network identifies sub-networks of pathways associated with keratinocyte differentiation, but it also expands the sub-networks to additional markers that have published evidence of interactive or regulatory mechanisms relevant for sub-networks characterization.
Conclusions: In conclusion, by combining longitudinal multi-omics data, we were able to detect temporal relationships and interactions between molecules over different omics layers. The integrative approach provided us with multiple novel insights into keratinocyte biology. This was particularly crucial for providing a comprehensive list of known and previously unrecognized major components of the epidermal reconstruction process.
References
1. Bodein, A., et al., timeOmics: an R package for longitudinal multi-omics data integration. Bioinformatics, 2021.
2. Bodein, A., et al., Interpretation of network-based integration from multi-omics longitudinal data. bioRxiv, 2020: p. 2020.11.02.365593.
3. Taylor, J.M., et al., Dynamic and physical clustering of gene expression during epidermal barrier formation in differentiating keratinocytes. PLoS One, 2009. 4(10): p. e7651.
4. Bachelor, M., et al., Transcriptional profiling of epidermal barrier formation in vitro. J Dermatol Sci, 2014. 73(3): p. 187-97.
5. Lopez-Pajares, V., et al., A LncRNA-MAF:MAFB transcription factor network regulates epidermal differentiation. Dev Cell, 2015. 32(6): p. 693-706.
Methods: To investigate the sequential events leading to epidermis differentiation at a multi-molecular level, we performed concurrent longitudinal high-throughput RNA-seq, miRNA-Seq, quantitative proteomics and metabolomics in a commercial reconstructed human epidermis model (SkinEthic Episkin Co, Lyon, France). In this regard, datasets were collected at different timepoints over twenty-eight days of culture. To identify complex relationships between omics layers, an analytic and integration framework was used for multi-omics longitudinal datasets that relies on multi-omics kinetic clustering using the timeOmics approach [1] and multi-layer network-based analysis using both data-driven and knowledge-driven building methods [2].
Results: Following data pre-processing, the molecules that varied most significantly during epidermis reconstruction were selected. Those with similar expression patterns over time were first modelled and clustered together. They showed distinctive temporal patterns characterized by their initial levels of abundance and variation during differentiation. For example, for molecules expressed on both gene and protein layers, temporal delay between mRNA formation and protein synthesis could be detected. In addition, enrichment analyses of GO terms performed on genes and proteins of each cluster recapitulated relevant biological functions previously described in skin reconstruction studies [3-5].
Next, by keeping the clustering information, a 4 layer network was built: two gene layers build from mRNA co-expression and miRNA-mRNA predictive regulations, a protein-protein interaction layer build from measured proteins and BioGRID known interactions, and a metabolite layer from KEGG pathways. Not only does this multi-layers network identifies sub-networks of pathways associated with keratinocyte differentiation, but it also expands the sub-networks to additional markers that have published evidence of interactive or regulatory mechanisms relevant for sub-networks characterization.
Conclusions: In conclusion, by combining longitudinal multi-omics data, we were able to detect temporal relationships and interactions between molecules over different omics layers. The integrative approach provided us with multiple novel insights into keratinocyte biology. This was particularly crucial for providing a comprehensive list of known and previously unrecognized major components of the epidermal reconstruction process.
References
1. Bodein, A., et al., timeOmics: an R package for longitudinal multi-omics data integration. Bioinformatics, 2021.
2. Bodein, A., et al., Interpretation of network-based integration from multi-omics longitudinal data. bioRxiv, 2020: p. 2020.11.02.365593.
3. Taylor, J.M., et al., Dynamic and physical clustering of gene expression during epidermal barrier formation in differentiating keratinocytes. PLoS One, 2009. 4(10): p. e7651.
4. Bachelor, M., et al., Transcriptional profiling of epidermal barrier formation in vitro. J Dermatol Sci, 2014. 73(3): p. 187-97.
5. Lopez-Pajares, V., et al., A LncRNA-MAF:MAFB transcription factor network regulates epidermal differentiation. Dev Cell, 2015. 32(6): p. 693-706.