Evolutionary Conservation and Divergence of Gene Coexpression Networks inGossypium(Cotton) Seeds

Abstract

The cotton genus (Gossypium) provides a superior system for the study of diversification, genome evolution, polyploidization, and human-mediated selection. To gain insight into phenotypic diversification in cotton seeds, we conducted coexpression network analysis of developing seeds from diploid and allopolyploid cotton species and explored network properties. Key network modules and functional associations were identified related to seed oil content and seed weight. We compared species-specific networks to reveal topological changes, including rewired edges and differentially coexpressed genes, associated with speciation, polyploidy, and cotton domestication. Network comparisons among species indicate that topologies are altered in addition to gene expression profiles, indicating that changes in transcriptomic coexpression relationships play a role in the developmental architecture of cotton seed development. The global network topology of allopolyploids, especially for domesticated G. hirsutum, resembles the network of the A-genome diploid more than that of the D-genome parent, despite its D-like phenotype in oil content. Expression modifications associated with allopolyploidy include coexpression level dominance and transgressive expression, suggesting that the transcriptomic architecture in polyploids is to some extent a modular combination of that of its progenitor genomes. Among allopolyploids, intermodular relationships are more preserved between two different wild allopolyploid species than they are between wild and domesticated forms of a cultivated cotton, and regulatory connections of oil synthesis-related pathways are denser and more closely clustered in domesticated vs. wild G. hirsutum. These results demonstrate substantial modification of genic coexpression under domestication. Our work demonstrates how network inference informs our understanding of the transcriptomic architecture of phenotypic variation associated with temporal scales ranging from thousands (domestication) to millions (speciation) of years, and by polyploidy.

Cis–trans controls and regulatory novelty accompanying allopolyploidization

“…Given that previous work in allopolyploids (e.g. wheat (Pfeifer et al ., ) and cotton (Hu et al ., )) was mainly based on aggregated co‐expression relationships of homoeologs, one future direction is to generate networks considering homoeolog expression separately, thereby allowing the direct evaluation of topological dynamics in terms of gain and loss of intra‐ and inter‐subgenome relationships (Conant & Wolfe, , ; Conant, ). Although co‐expression relationships do not necessarily represent physical interactions between cis and trans regulatory elements, the gene‐to‐gene interconnections that are inferred based on the ‘guilt‐by‐association’ principle provide an alternative and parallel approach for understanding the impact of genomic merger and doubling, under the same analytical framework used for genes outside of a network context.…”

Genetic analysis of the transition from wild to domesticated cotton (G. hirsutumL.)

“…Interestingly, there appears to be little bias toward differential expression of genes under fiber-related QTL versus non-fiber QTL for these fiber-derived expression data. This may reflect a general overlap between fiber-relevant genes (e.g., cell wall, cytoskeletal genes, etc) and those involved in broad plant phenotypes, as well as the remarkable increase in gene coregulation during domestication (Hu et al 2016). Therefore, while we note differences in DGE for possible candidate genes from any trait category, the relevance of this fiberderived DGE to non-fiber traits is unclear.…”

“…The complex interactions among genes important to understanding the QTL recovered remain to be elucidated, but many important enabling tools for such analyses have been developed. For example, gene coexpression network analyses can reveal modules of interconnected genes involved in key traits, as shown for cottonseed (Hu et al 2016) and fiber (Gallagher et al in prep), using the comparative context of wild versus domesticated G. hirsutum. In these examples, domestication appears to have increased the coordinated expression among genes and gene modules relevant to domesticated phenotypes.…”

Homoeologous gene expression and co-expression network analyses and evolutionary inference in allopolyploids

“…Finally, as previously described [15], we compiled a list of genes that are overrepresented with the gene-to-gene paired DC relationships (see above) to identify differentially co-expressed genes (DC genes). Briefly, the probability follows the binomial distribution model:…”

“…These responses are many and varied, including biased homoeolog expression, conditionspecific differential homoeolog usage, transgressive expression levels, and expression level dominance. Duplicated gene expression patterns are coordinated in ways that are not fully understood and which depend on myriad factors, including dosage effects, gene balance, interactions among divergent cis and trans-acting factors, and various topological aspects of gene networks [6,[15][16][17][18].…”