Plants possess a remarkable ability to coordinate the timing of developmental events. Dormancy, bud burst, seed germination, and senescence are just a few of the phenological processes that rely on a close coordination with climatic cues. In the case of annuals, flowering time is an especially critical life-history trait, one that is shaped by evolution to maximize fitness, fecundity, and reproductive success (Amasino, 2010). Studying the physiological and molecular mechanisms that underlie these phenological processes has kept the plant biologist busy for decades. Such investigations show that plants rely on an intricate signaling pathway to fine-tune the seasonal and developmental timing of flowering. Foremost among the environmental cues are light and temperature, and their interaction with a range of complex molecular mechanisms. To date, several major genetic pathways and a number of key regulatory genes are implicated in the control of flowering. In this issue of New Phytologist, Chew et al. (pp. 654–665) build upon this wealth of information and present a fascinating series of observations that further tease apart the role of light and temperature in the control of flowering for the model organism Arabidopsis. The research integrates theory, molecular biology, and field data in a way that epitomizes the iterative nature of data collection, model evaluation, and the generation of new hypotheses. This paradigm is often invoked, but seldom achieved to the degree reflected by these investigators. As Chew et al. illustrate, models can facilitate not only the understanding of complex behavior in plants, but also allow new insights to be gained that can in turn be applied across molecular and whole plant scales.
‘The research integrates theory, molecular biology, and field data in a way that epitomizes the iterative nature of data collection, model evaluation, and the generation of new hypotheses.’
Molecular mechanisms and predictive modeling of system behavior
The previous decade has seen numerous investigations that together provide remarkable insight into the molecular mechanisms responsible for flowering time as shaped by seasonal and developmental cues. An increasingly apparent outcome from these studies is that the pathways driving flowering are highly diverse and interactive. A key breakthrough in flowering was the demonstration that CONSTANS (CO ) and GIGANTEA (GI ) are necessary for activation of FLOWERING LOCUS T (FT), which we now know to be florigen, once considered the ‘Holy Grail’ by the flowering research community (Turck et al., 2008; Amasino, 2010). This simple network of interacting members continues to be layered with additional complexity. It was found, for example, that flowering can occur in Arabidopsis not only under long day-length, but also for short days when the gibberellin (GA) pathway is induced. Environmental interactions add to this complexity. Dense vegetative shade was found to influence flowering through the involvement of phytochrome photoreceptors with phytochrome-A promoting flowering by stabilizing the CO protein while phytochrome-B suppresses flowering by preventing CO accumulation (Franklin & Quail, 2010). Additional interactions became apparent with the observation that prolonged exposure to cold temperatures (e.g. vernalization) promotes flowering and that genetic variation in FRIGIDA (FRI) and FLOWERING LOCUS C (FLC) alleles are associated with vernalization requirements (Amasino, 2010).
While the last decade can be characterized as providing key advancements to our understanding of the numerous interacting molecular mechanisms contributing to flowering, the next decade of flowering research may well be characterized as the modeling of those relevant mechanisms to address questions of ecological and agronomic importance. Flowering is a key component of phenology and is considered a major trait when assessing the consequences of climatic change on species distribution and ecosystem productivity (Cleland et al., 2007). It is in this light, that the modeling endeavors of Chew et al. truly become apparent. Their approach was built upon the accomplishments of earlier combined temperature and photoperiod models, and that of Wilczek et al. (2009) who included genetic and molecular information on pathway structure and ensuing physiological responses. The model is traditional in the sense that flowering is predicted based on the threshold of accumulated photothermal developmental units, however, novel network-based scaling factors are used as proximal representations of specific gene activities in the regulatory pathways. The scaling factors were determined through model optimization using field data, where the value for each specific genotype was constrained based on its loss or gain of function relative to wild-type plants. Flowering times of seven genotypes (two in the Ler background and five in Col) were used in the model optimization, and the model was later extended to describe the field data of phyA and phyB mutants. This combination of field data collection and model parameterization allowed predictions of flowering to be adjusted according to environmental sensitivity specific to genotype, thereby providing a major step forward in advancing traditional phenology models to include genetic information.
Integration of data-driven model development, adjustment and evaluation is rarely exercised in molecular biological studies and thus model-aided hypothesis generation and testing can be hampered. This is certainly not the case with Chew et al., who articulately demonstrated that deviation of observed from predicted values can be quite insightful biologically. This was initially shown when they found their original model able to predict flowering of spring and summer cohorts reasonably well, but not autumn cohorts. Resolution of this discrepancy was achieved by incorporating night temperature into their revised model, which improved model prediction. Inclusion of night temperature may isolate the known sensitivities of GA to day/night temperature; thereby explaining, in part, the predominant role that GA floral regulation plays during shortened photoperiods. Further insight was garnered when the model was tested on data from phytochrome mutants (Fig. 1), where optimization of photoperiod parameters sufficed to predict phyA null mutant flowering but not phyB. In fact, accurate fit to the phyB data were only achieved when temperature adjustment was removed, suggesting a possible mechanism for the role that phyB plays in both temperature and photoperiod perception. The authors note that this may explain the paradoxical observation that phyB overexpressors and phyB null mutants both flower early compared to wild types.
Enabling technology and resources
Identifying the genetic and environmental controls on flowering time has benefited, as shown by Chew et al., from access to mutants obtained through forward and reverse genetic manipulations. Breakthroughs have also been facilitated by the increased availability of mapping populations for quantitative trait loci (QTL) analysis and the collection of plants from across diverse climates and their use in genome-wide association studies. Weigel (2012) recently emphasized that resources like these were a valuable component of the genetic tool kit for the molecular dissection of heritable traits, including intra-specific variation in flowering time. Along these lines, Li et al. (2010) examined flowering time across a diverse collection of 473 Arabidopsis accessions, totaling > 12 000 plants. These authors used specially-programmed growth chambers to impose environments that mimicked prevailing climates at multiple geographic locations (Fig. 2). Increasingly sophisticated genetic resources coupled with growth and imaging capabilities (Zhang et al., 2012) provide the basis for in-depth scientific investigation, model development, and testing. Add to this the emerging fact that genotyping of very large numbers of plants has become both tractable and affordable with next-generation sequencing methods, will only further facilitate linking genotype and phenotype.
Investigations in the years to come will undoubtedly refine our understanding of temperature and light in flowering. This remains an active area of investigation. In addition, and as shown so convincingly by Chew et al., genetically-informed models will be an increasingly important component of this research and provide opportunities to bridge laboratory and field studies. In fairness, the application of mathematical modeling to flowering time is not new, as various strategies including neural networks (Welch et al., 2003) have been used to interpret complex circuit behavior in the control of flowering for a decade. It is, however, not the use of models per se but rather the way models were used in the Chew et al. investigation that deserves special recognition. Interestingly, Penfield (2008, p. 625) in his New Phytologist Tansley review on temperature perception and signal transduction in plants wrote ‘In the next decade we will develop computational models of these signaling processes and this will enable comprehensive approaches to hypothesis generation and testing’. It appears that Chew et al. and others (Wilczek et al., 2009) have already made strides in this direction and, as a result, have significantly compressed that timeline.
Support provided by the US Department of Energy, Office of Science, Biological and Environmental Research Program. Oak Ridge National Laboratory is managed by UT-Battelle, LLC, for the US Department of Energy under contract DE-AC05-00OR22725.