Identification of Key Regulators of Guard Cell Functions Using Systems Biology Approaches

Open Access
Li, Song
Graduate Program:
Integrative Biosciences
Doctor of Philosophy
Document Type:
Date of Defense:
July 19, 2010
Committee Members:
  • Sarah Mary Assmann, Dissertation Advisor
  • Sarah Mary Assmann, Committee Chair
  • Reka Z Albert, Committee Chair
  • Andrew George Stephenson, Committee Member
  • Claude Walker Depamphilis, Committee Member
  • Debashis Ghosh, Committee Member
  • systems biology
  • guard cells
Over the past decade, fully sequenced genomes and the advent of the high throughput technologies such as transcriptome profiling with microarrays, have provided biologists catalogs of the molecular components of many biological systems. While traditional genetic and molecular biology approaches focus on identifying functions of one individual component at a time, systems biology approaches have become essential for biologists to understand the complex interactions of the cellular components, and to distill key regulators out of the rich data that are generated by high throughput experiments. I have undertaken two projects that address two general questions in systems biology: 1) how to reconstruct the network of the interacting molecular components, and to simulate the cellular process that is carried out by the molecular network. 2) Given a compendium of gene expression data from multiple tissues in a multicellular organism, how to identify novel genes that function in a specific cell type. In both projects, I have used plant guard cells as a model system, while each project has been developed using specific computational methods. Dozens of cellular components have been found to be involved in the guard cell ABA response, however, most interactions between these components are indirect, and the detailed parameters of known direct reactions are largely unknown. A Boolean network approach is adopted to address the problem of lack of details regarding the guard cell signal transduction network. Random asynchronous update and random initial states of the components are introduced to model both the uncertainty that is inherent in the intracellular signaling process as well as our incomplete knowledge of the signaling process. Simulations are carried out to determine whether knockout of each component has an impact on the ABA induced stomatal closure. Computational predictions are validated by wet-bench experiments using chemical inhibitors that “clamp” intracellular signaling components. In a multicellular organism, each cell type performs specialized functions using genes that are expressed in that cell type. To understand the general principle of how genes are expressed in multicellular organisms, tissue-based gene expression patterns are analyzed for thousands of genes from four multicellular organisms: human, mouse, rice and Arabidopsis. Expression kurtosis is found to be the common organizing principle of tissue preferential expression in all four organisms, thus is used to define in which tissues a high kurtosis gene is preferentially expressed, for every high kurtosis genes in both human and Arabidopsis. To identify novel genes that are essential to guard cells’ functions, transcriptome data are obtained using Arabidopsis transcriptome microarray. A linear model approach is used to merge the expression profiles generated by our group with published results from other published transcriptome data before the identification of high kurtosis genes that are preferentially expressed in guard cells. Many high kurtosis genes are found to be preferentially expressed in multiple tissues. Using high kurtosis genes as signatures for tissue functions gene-centric tissue networks are constructed to explore the similarities between different tissues. Tissues from similar organ origination are found to group into modules in the gene-centric tissue networks, supporting the predictive power of the tissue network approach. Four genes that are newly predicted to function in guard cells and/or roots are validated using knockout mutants, and the phenotypes of mutants agree well with the in silico predictions. The results of gene tissue associations and the tissue networks are of general interest to biologists. A web interface that allows interactive visualization and exploration of the gene centric tissue networks is available at