Chasing Scattered Genes: Identifying Specialized Metabolite Pathway Genes through Global Coexpression Analysis[OPEN]

Plants produce scores of specialized metabolites (SMs) to attract or repel the organisms around them and to cope with life in a variable environment. For thousands of years, we have beenexploitingthesecompoundstofeed,heal, and anoint us. Many more SMs remain to be discovered: The chemical constituents of only 15%of theestimated350,000plantspecieson Earth have thus far been explored (Wurtzel and Kutchan, 2016). Since SMs are generally not required for plant growth or reproduction, the underlyinggenesandpathways leading to their production have diversified greatly over time and are not well conserved among species, making them difficult to identify through standard homology searches. However, genes withinanSMpathwaycanbeidentifiedthrough their shared regulatory network, since the successful production of an SM requires the underlyinggenes tobeexpressedat the right time and place. Searches for coexpressed genes from global gene expression data have shed light on SM pathways in various plants, but technical constraints and limited data have hampered such analyses. In addition to their tight regulation, genes in SM pathways, at least in bacteria and fungi, are often clustered together in the genome, forming biosynthetic gene clusters (BGCs). Powerful tools are used to identify BGCs and to predict their involvement in SM pathways. While most known plant SM pathway genes are dispersed across the genome, several plant BGCs have been identified andmany more have been predicted (Schlapfer et al., 2017), and the idea thatSMpathwaygenes in plants tend to beclustered together hasbeen gaining traction. If this is not the case, however, techniques for finding plant SM genes based on chromosomal proximity, an easyto-detect feature, would fail to uncover most SM pathways, prompting Wisecaver et al. (2017) to investigate this issue using data from eight model plant species. Based on the assumption that genes in an SM pathway form tightly associated coexpression modules, the authors used pairwise measurements of gene coexpression data from hundreds to thousands of experiments to construct mutual rank (MR)-based coexpressionnetworks.Geneswerethenassigned to modules of tightly coexpressed genes using the ClusterONE tool. Focusing on small (<50gene)modulestoreflect thetypicalsizeof an SM pathway, the authors looked for modulescontainingSMpathwaygenesinthePfam database, finding the fewest suchmodules in the green alga, Chlamydomonas reinhardtii, and the most in the mustard, Brassica rapa. Manymodules (;15 to just over 50%) contained two or more known SM biosynthetic genesandgenesenriched inSM-related functional categories, as well as many experimentally validatedSMpathways. For example, this analysis identified almost all genes involved in themethionine-derived aliphatic glucosinolate (metGSL) biosynthesis pathway and associated biochemical processes in Arabidopsis thaliana (see figure). This approach also revealed all six functionally characterized SM pathways known to form BGCs in the eight plant genomesexamined.However, anexamination of predicted but not experimentally validated BGCs suggested that these clustered genes are not coexpressed and do not form coexpression modules and might thereforenotcorrespondtofunctionalSMpathways after all. Thus, proximitymight not be a reliable index for identifying SMpathways, sincemost are likely scattered, not clustered. Instead, SM pathwaysmanage toproduce their highlycoveted products through coordinated expression of their genes, a trait that can now be exploited to facilitate their discovery.