Final Report Summary - TIMET (Linking the clock to metabolism)
Executive Summary:
The ambition and scope of TiMet was unprecedented in the field of plant systems biology. The overall aim was to understand the growth rate of plants quantitatively, and in particular, how the circadian clock controlled key metabolic pathways to change plant growth rate over the day-night cycle and across the seasons. TiMet participants combined the methods and insights from multiple disciplines, from studies of enzyme activity in metabolism, to genome-scale molecular biology and plant physiology, and from statistical to mechanistic and multi-scale modelling approaches.
A test case for plant systems biology
Three, closely-linked biological systems were selected for the project. The interconnected genes of the 24-hour biological clock were already a paradigm for systems biology. The pathways of starch metabolism provide carbon storage that is essential to allow plant growth at night. Metabolic products of the isoprenoid pathway include chlorophyll and key plant hormones. The key achievement was to move from disjoint understanding of these separate systems in a static context, to dynamic, systems-level understanding linked quantitatively to the growth of the whole plant.
New results, methods, models and understanding
1. The circadian clock.
TiMet’s models provided a new framework to understand the dynamics of clock gene expression, integrating key clock genes and linking to multiple clock-output pathways for the first time in any organism. Transcriptome and proteome data showed how daily rhythms of molecular activity are affected by changing photoperiods, metabolic and clock mutants, and in climate change scenarios.
2. The starch metabolic pathway.
Molecular mechanisms that adjust the rates of starch synthesis and degradation under different photoperiods were discovered. Unexpectedly, starch degradation continues during the day. TiMet’s results explain the regulation of starch on an hourly timescale, and point to system-level feedback from growth, whereas previous understanding contrasted two, uniform states for day and night.
3. The isoprenoid metabolic pathway.
Rhythmic regulation of the isoprenoid pathway was characterised at the transcriptional and post-translational levels, identifying key regulators in the chloroplast, and integrated into the first mathematical model of this system.
4. Plant growth.
The project documented for the first time the remarkable flexibility in the partitioning of plant growth between day and night, revealed how energy and carbon utilisation are optimised, and quantified how global factors control growth rate. In our multi-scale Framework Model, a simulated clock mutant explained the mutant plant’s biomass, representing the first time that altered gene expression dynamics have been linked quantitatively to such a high-level property in any organism.
5. Extensions to unexpected areas.
TiMet’s analysis of plant growth was extended to natural conditions, to understand natural genetic variation within and across species, and formulated as a problem of micro-economic optimisation.
In addition to the many particular results, TiMet research validated three new and general operating principles for biological timing. Starch degradation implements “arithmetic division” to set the upper limit on the nightly starch degradation rate. The isoprenoid protein DXS revealed “anti-rhythmic” regulation, where transcriptional and post-translational rhythms in opposite phases combined to maintain a constant protein level. “Translational coincidence” uses daily rhythms in RNA level to control the abundance of hundreds of proteins, not within the day but across the seasons.
Project Context and Objectives:
Photosynthesis provides all of our food, and a significant proportion of our fuel and industrial raw materials. Demand for the products of photosynthesis will increase dramatically in the next few decades. At current rates of crop yield improvement, the FAO predicts that the world will face a major shortfall (ca. 30%) in the supply of primary foodstuffs relative to demand by mid-century. Optimising photosynthesis is a promising route to increase crop yield potential and a current target for research ranging from ecology to synthetic biology. However, the Earth’s rotation predictably removes sunlight – and hence photosynthesis – for a significant part of each day, and each plant’s lifespan. Plants must orchestrate the accumulation and utilisation of photosynthetic products over the daily cycle to avoid periods of starvation, and thus optimise growth rates. The many other plant processes that are directly or indirectly affected by the day/night cycle must be regulated in parallel, to avoid conflicting requirements and to optimise plant productivity. The TiMet project aimed to understand how plants achieve this daily optimisation.
Project context
Before the TiMet project, we and colleagues had shown:
(i) that plants possess a circadian clock, comprising a circuit of regulatory genes that provide predictive, temporal regulation of many plant processes over the day/night cycle,
(ii) that accumulation of starch provides a store of the sugar derived from photosynthesis that can be used at night, providing both cellular energy and a substrate for metabolic pathways such as isoprenoid biosynthesis that are important for growth.
(iii) that these systems are intimately interlinked. The circadian clock controls the rate of starch utilisation at night, so that this store lasts for the entire night, both in long, winter nights and in short, summer nights.
The loss of any one of these systems was known to result in serious decrease in plant productivity, so these biological processes were central to agricultural productivity. Importantly, the concentration of understanding, data, tools and resources available in the laboratory model plant Arabidopsis thaliana (and in this species only) made detailed studies of these processes experimentally tractable. The combined system was therefore a test case, in which to elucidate the coordination of responses of multi-level biological networks to incoming environmental signals. The was a strong case that starch metabolism might be subject to particular regulation that was not representative of other pathways, hence we included the isoprenoid pathway, which produces critical components for growth, from key cofactors such as chlorophyll, to antioxidants and plant hormones. We anticipated that these studies would require mathematical modelling to understand the complex, dynamic responses in gene regulation, metabolism and growth, and to link them across different time and length scales. However, the dynamics of a gene regulatory network had never been linked mechanistically, via altered metabolism, up to the level of whole-organism biomass, in any multicellular species.
At the start of the project, our understanding was disjointed and at disparate temporal resolution. The “chronobiology” and “metabolism” research fields were traditionally rather separate but now needed to be combined, in order to discover the mechanisms by which the clock exercises control over metabolism. Biochemical studies of detailed metabolic processes were also traditionally separated, both among areas of metabolism and from understanding of macroscopic growth. They too needed to be linked, in order to understand in depth how alterations of metabolism inhibited growth, in quantitative rather than qualitative terms.
These limitations were reflected both in experimental data and in the mathematical models available to formally represent our understanding of these systems, from (i) the operation of the circadian clock and its environmental light input, to (ii) the control of downstream clock outputs, to (iii) their joint effects on starch metabolism and in other downstream central metabolic pathways like isoprenoid metabolism, to (iv) the impact on whole plant growth. The dynamics of the circadian clock were studied at hourly resolution and were a paradigm for genetic responses to environmental signals. The daily, rhythmic regulation and negative feedback among the clock genes were also a paradigm for modelling of gene networks. The transcriptional regulators expressed at dawn LHY and CCA1 were well known, along with a family of PRR genes including TOC1, which regulated LHY and CCA1 by an unknown mechanism (they are now known to be direct transcriptional regulators). Additional components such as GI were known to link the clock with the pathway that accelerated flowering under long photoperiods. However, some of the most important clock components known could not be integrated into the models. In particular, mutants of ELF3, ELF4 and LUX genes had very severe effects up to abolishing circadian timing, but their biochemical functions were unknown (they are now known to form a repressor termed the Evening Complex). Models of the clock gene network had not been linked to any clock output, for any organism.
Starch metabolism in plants was studied at daily resolution, discriminating between two regimes. In the day, starch was thought to be synthesised but not degraded; in the night, starch synthesis ended and the degradation pathway mobilised sugars for use in the cell. Models of starch metabolism in other plant species had either been linked to photosynthetic inputs, with detailed biochemical mechanisms but without outputs to growth, or had been linked to growth but in a phenomenological context, without biochemical mechanisms. No dynamic models existed for isoprenoid metabolism. The only growth models with molecular detail focussed on developmental processes leading to shape alone, without any biomass, or accounted for biomass without molecular mechanisms. Within the modelling communities, research groups relied routinely on a subset of methods, rather than combining the multiple methods required at different stages of a systems biology investigation.
The TiMet project therefore assembled world leaders in experimental and theoretical plant systems biology, to understand the regulatory interactions between the clock gene circuit and plant metabolism, and their emergent effects on whole-plant growth. Many of the underpinning methods and resources for our studies of the Arabidopsis clock, starch and isoprenoid pathways were originally created by the TiMet partners. TiMet combined their exceptional and often world-leading, experimental facilities in order to address the problem at multiple levels of biological organisation, both in detailed assays and transcriptome- or proteome-wide. They were applied to analyse the responses to day-length and light-quality regimes that modify and disturb clock function in wild-type plants and in a large collection of mutant lines deficient in the clock, starch and isoprenoid metabolism. The data were organised in new infrastructure, and informed innovative modelling platforms that will permit functional analyses, generated further insights and testable hypotheses, which were then verified by further experiments.
Project objectives
Our objectives were to:
1. Establish a single experimental system to study clock, starch and isoprenoid subsystems, with mechanistic, mathematical models of each component, parameterised for this system.
2. Use high throughput, cutting-edge ‘omics platforms to acquire network-wide and targeted data on the network’s dynamic responses over the day/night cycle, with appropriate data handling to support data sharing and integration and binding to models.
3. Manipulate the network to discover the mechanistic connections from the clock to starch metabolism and isoprenoid synthesis, as well as inputs from these metabolic pathways that control the clock.
4. Include the new connections in the mathematical models of the component pathways.
5. Analyse both models and data in order to understand and ultimately predict the effect of each connection on the emergent properties of the intact system, including the carbon and energy available for long-term plant growth.
6. Test this understanding through dynamic manipulation of particular subsystems, monitoring outcomes at both the cellular and whole-plant levels.
7. Validate and refine the integrated model through iterative rounds of experimentation. Predict experimental strategies to further dissect the molecular network and to engineer/select desirable whole-plant behaviours, corresponding to crop traits.
The project aimed to develop fundamental understanding of this important system, as a test case for using model-guided, mechanistic tools to analyse biological regulation across scales. This was only possible in Arabidopsis. Establishing the toolset and proof-of-principle in this species, however, will now justify and direct future work to apply this systems biology approach to other biological processes, and to other species including the best-known crops.
Project Results:
The TiMet project achieved systems-level understanding of the links between circadian timing, metabolism and plant growth, crossing scales from intracellular pathways to the whole organism, and at a new level of temporal resolution. Over the 5-year lifespan of the project, the consortium has reached 37 Milestones and provided 33 Deliverables to the Commission. The consortium has published 52 research articles directly supported by TiMet, with 2 further papers in press, 7 book chapters or sections and 1 paper in conference proceedings. Many of these studies have appeared in the leading, international, peer-reviewed journals including Science, PNAS, Molecular Systems Biology, eLife, Plant Cell and influential review journals such as Annual Reviews of Plant Biology. In addition, 17 further publications are in preparation or under review for similar journals.
TiMet was a highly-integrated project, by design. Separating the results for presentation in any linear narrative necessarily results in multiple discontinuities. We therefore integrate the results from multiple work-packages into thematic sections below, in order to minimise these discontinuities, and note the relevant workpackages in the section headings. The close links that remain between the sections are indicated in the text. Moreover, new research methods were required in several cases to achieve the project’s ambitious aims. The development of these methods is usually described in the relevant results sections below or, for data infrastructures and one modelling method, in the final section.
1. New understanding of the biological clock and its pervasive effects.
TiMet aimed to develop a mathematical model of the clock genes, as they operated in the mature rosette leaves used in studies of starch metabolism rather than in the young seedlings typically used in studies of the circadian clock.
a. The circuit of genes that creates 24-hour biological rhythms (WP1, 2.4)
The existing, partial “P2010” model was first updated to represent newly-published findings. The key clock components ELF3, ELF4 and LUX were added, forming the “P2011” model, which also predicted a change in the function of the TOC1 protein. The prediction was verified experimentally by an independent group and by collaborators, who also showed that TOC1 controlled a greater than anticipated number of the other clock genes. An updated “P2012” model was then developed to include these data, explaining for the first time the clock gene expression pattern of the toc1 mutant. A stochastic version of the P2010 model showed how noise due to biological variability could affect the rhythmic gene expression.
To complement the subjective, expert judgement involved in building these models, statistical methods were also developed to learn the circuit connections that link the clock genes. This was a challenge for machine learning. Progress was aided by using simulated data from the stochastic P2010 model as a test case, because the true circuit underlying those data was completely known. Revised connections have now been suggested by these results.
Significant results:
• The P2011 model correctly predicted that TOC1 is a repressor, not an activator, of gene expression.
•
• Representing ELF3, ELF4 and LUX in P2011 changed our view of the clock’s operation.
•
• TiMet’s Hierarchical Bayesian Regression (HBR) learning method outperformed all competitors.
•
• An ensemble method, combining HBR and 14 others methods, outperformed all single methods.
•
b. Effects of environmental and genetic variation (WP1, 2.5)
The clock genes’ RNA profiles were measured in soil-grown plants of wild type (Col0 and Ws2) and circadian clock mutants (gi, toc1, prr7 prr9, lhy cca1 and elf3), and compared to data of collaborating researchers from seedlings that were grown with or without exogenous sucrose. Plants with inhibited or enhanced isoprenoid metabolism were also tested. The RNA profiles of the wild-type clock genes varied very little in these different conditions. The detailed behaviours of some clock mutants, however, suggested new connections among the clock genes. The absolute levels of clock gene RNA were measured for the first time. The wide range of expression levels was surprising, considering that all the clock genes are believed to function together in the same circuit.
Measuring the clock RNA profiles in plants grown with 4, 6, 8, 12 or 18h of light per day revealed how the change in day length progressively affected the clock’s timing. The “double-dawn” light conditions that were used in later studies (see Section 2.c.iii) were studied in greater detail, using our luciferase (LUC) reporter imaging facilities. These in vivo bioluminescence assays offer high temporal resolution and avoid the biological variation due to destructive sampling of different plants at each time-point for RNA. As little as 1µmol/m2/sec of light fluence rate at dawn was sufficient to set the phase of the rhythmic reporters. The LUC imaging assays also allowed us to test whether biological timing was uniform across the leaf. Leaves grown without environmental cycles for up to three weeks showed waves of rhythmic clock gene expression crossing the leaf, forming various spatial patterns. However, light-dark cycles fully synchronized all parts of the leaf, suggesting that the spatial patterns rarely occur in nature.
Arabidopsis Col-0 and two further accessions were also tested under five different temperature regimes, designed to investigate the impact of night temperature on metabolism and growth (see Section 2.a).
Significant results:
• Clock gene expression profiles were tested under the most diverse set of conditions ever reported.
•
• The first absolute quantification showed just ~50 PRR9 RNA copies/cell at the peak, the lowest level of the clock genes; the highest was LHY RNA, at over 1500 copies/cell.
•
• The clock model need not be specially adapted for rosette leaves, because clock gene expression patterns were very similar in wild-type tissue under all conditions, including with added sucrose.
•
• Almost all the progressive changes in phase under different photoperiods were matched by the existing clock models, indicating that these models are sufficient to analyse our metabolic data.
•
• The plant clock responded both to low intensity, photosynthetically-inactive light and to high-intensity light: their joint effects located subjective dawn to an intermediate time.
•
• Under constant light, different regions of a single leaf can differ in clock phase by up to 17h.
•
• Spatial patterns of clock phase indicate that clocks in neighboring regions are weakly coupled.
•
c. Testing for widespread, rhythmic regulation of mRNA and protein levels (WP1, 2.3 2.5)
The impact of the circadian clock on the broadest possible set of Arabidopsis mRNAs and proteins (the transcriptome and proteome) were analyzed in the wild type and clock mutants described above, and across multiple photoperiods, at dusk (end-of-day; EOD) and dawn (end-of-night; EON) time points. The disruption of core circadian clock components has a major impact on the transcriptome Changes in RNA transcript levels are not always reflected at the protein level. Thus, mutation of the clock genes does not globally affect the proteome but rather selectively influences specific protein targets.
Significant results:
• Altering the photoperiod from 6h to 18h has a profound effect: more mRNAs change in level at dawn in this comparison than change between dawn and dusk in the standard, 12h photoperiod.
•
• Fewer of the detectable proteins change expression between dawn and dusk, as expected, but the number changing in our results is greater than published estimates.
•
• Genes associated with the starvation (carbohydrate) response and encoding transcription factors (TFs) and protein kinases changed RNA level the most in the clock mutants. At the protein level, most changes were observed for carbohydrate response, phenylpropanoid, TFs and isoprenoid related genes.
•
• Proteins with a higher abundance in longer photoperiods included enzymes involved in the biosynthesis of glucosinolates, auxin and jasmonate hormones, and the starch degradation pathway. Proteins in the RNA processing pathway, lignin biosynthesis as well as in plastidial ribosomal proteins decreased in abundance.
•
d. Effects of daily timing on central metabolites (WP1, 2.1 2.5)
To test the effects of the clock and the altered conditions upon plant metabolism, photosynthesis, respiration, starch and key soluble metabolites were analyzed in the complete timeseries from wild-type plants, compared to clock mutants, altered temperature regimes and altered photoperiods, using the samples tested for the clock gene mRNAs (see 1.b above). These results provided the information to interpret starch metabolism in the context both of the carbon supply from photosynthesis and of its utilization in respiration and growth. Complementary metabolic labelling methods were also used to investigate the fluxes of carbon to protein synthesis, the other chemical constituents of biomass and the distribution among organs in whole plants (see part 4, below).
Significant results:
• Plants in photoperiods up to 12h were sugar-limited (“source limited”); plants in 18h photoperiod had excess sugar and were instead limited by the “sinks” that consume the sugar during growth.
•
• A previously undiscovered, circadian rhythm was detected in the level of a key sugar-phosphate.
•
• The lhy cca1 and prr7 prr9 mutants had the widest, and often reciprocal, impact on metabolite levels including sugar, amino and organic acid levels.
•
• Expert analysis of the entire data set suggested that distinct outputs from the clock affect starch degradation (controlled by multiple clock components; see part 2 below), organic acid and amino acid metabolism (controlled by morning clock components; see part 4), and reducing sugars (controlled by the evening complex genes).
•
• The machine learning ensemble method, including possible delays between RNA levels and the effect on metabolites, independently suggested results with a striking overlap: three clock genes connected to components of the starch synthesis pathway; one of the morning clock components was linked to levels of organic and amino acids; the suggested link from the evening complex was, however, to the sugar-phosphate rather than the unmodified sugar.
•
The striking results on starch metabolism are described in greater depth in section 2, with detailed analysis of the isoprenoid pathway in section 3, and the links from the clock and metabolism to growth in section 4, below.
2. New understanding of starch metabolism
Starch metabolism is critical for growth in Arabidopsis and many other plant species, because starch provides the major source of sugar to maintain cell function when darkness prevents C3 photosynthesis.
a. Effects of clock mutations, photoperiod and temperature (WP1.1 1.2)
The results from the clock mutants, altered photoperiod and temperature regimes (section 1b) on the timing of starch synthesis and degradation extended and reinforced the earlier results that prompted our studies of the starch pathway. Starch is synthesised at a higher rate in short photoperiod cycles, and more of the available photo-assimilate is partitioned into starch, compared to longer photoperiods. Starch degradation proceeds at a nearly constant rate from dusk to dawn, controlled to consume all but 10% (often as little as 5%) of the starch at dusk. The high-quality TiMet data provided the metabolic and growth context to understand the response of starch metabolism to many different manipulations.
Significant results:
• The clock-dependent timing of starch degradation led to stable but photoperiod-dependent sucrose levels across most of the night.
•
• Starch synthesis begins rapidly in the day, except in the 18h photoperiod, when there is a plateau for a few hours after dawn before starch levels start to rise.
•
• In 18h photoperiods, another plateau in starch synthesis occurs before dusk, when starch levels cease rising at the end of the day.
•
• Starch synthesis rates increased reliably under short photoperiods.
•
• Starch degradation rate was affected by photoperiod and clock mutants.
•
• Surprisingly, the starch degradation rate was hardly affected by night-time temperature.
•
• In sink-limited plants, however, starch degradation rate became temperature-sensitive.
•
The first conclusion was that Arabidopsis has a conservative strategy, storing more carbon as starch during short photoperiods in order to eke out these reserves during the following, long night. However, TiMet’s series of experiments (below) revealed considerable flexibility in some parts of this programme, and identified several of the mechanisms involved.
b. Control of starch synthesis (WP2.2)
i. Potential links from the clock
Both starch synthesis and degradation rates were predictably and reciprocally affected by photoperiod. Data on the circadian clock mutants tested the role of the clock in this photoperiod-dependent response.
Significant results:
• Most of the clock mutants that altered starch degradation rates (see below) did not affect the increase in starch synthesis rate in short photoperiods. Their pleiotropic effects, including starch excess phenotypes, might complicate these effects.
•
• In the gi and fkf1 mutants, however, starch synthesis did not respond to photoperiod. GI and FKF1 proteins form a complex essential for the clock-controlled photoperiod response of flowering time (see Section 4b).
•
The response of starch synthesis to changing day length is probably not directly controlled by the circadian clock components. The clock may still be indirectly involved, perhaps through a process involving GI and FKF1, which appear to form a novel link between the clock and primary metabolism. Both of these genes are strongly clock-regulated, with peak expression in the mid to late day. As the origin of the signalling mechanism from the clock was recalcitrant, the project focussed on identifying the target enzymes in the synthesis pathway that mediate the response.
ii. Target enzymes in starch synthesis
We hypothesised that photoperiod control of starch synthesis might be mediated by modulation of the activity of ADPglucose pyrophosphorylase (AGPase), an enzyme with several potentially important regulatory properties. Substituting altered forms of AGPase confirmed this hypothesis.
Significant results:
• The photoperiod response in starch synthesis was absent from plants expressing an unregulated, bacterial form of AGPase, though their starch profiles were normal in 12h photoperiods.
•
• The general pattern of starch accumulation in plants expressing a mutated, redox-insensitive, constitutively-active plant AGPase was similar to that of wild type-plants.
•
• Adjustment of the rate of starch synthesis to changing day lengths is not brought about by modulation of subunit composition, as mutants lacking individual subunits could still adjust starch synthesis to changing day lengths, and the immune-precipitated enzyme composition was the same in both photoperiods.
•
The regulatory properties of the plant AGPase are therefore essential for the response of starch synthesis to photoperiod but the redox modulation of AGPase activity was not necessary. The remaining well-understood regulatory property of AGPase is its allosteric activation and inhibition by 3-phosphoglycerate (3PGA) and inorganic phosphate (Pi) respectively. This property is believed to be important for the adjustment of the rate of starch synthesis in response to imbalances between sucrose synthesis/demand and carbon assimilation within a single day. TiMet’s results extend the role of allosteric regulation to the photoperiod-dependent increase in synthesis rate, which occurs without a major imbalance in synthesis and demand for sugar. This adjustment might reflect altered daily patterns of growth and hence demand for photosynthate in different day lengths, rather than a direct influence of the clock on the synthesis pathway itself, and is therefore discussed further in Section 4. This conclusion contrasts with more direct clock regulation of starch degradation.
c. Adjustment of starch degradation (WP2.1 2.2 2.5)
i. Potential links from the clock to target enzymes in the pathway
TiMet extended earlier work on the clock mutants. The strong, short- or long-period mutants clearly degraded their starch reserves to expire too early or late, respectively, relative to the actual time of dawn. The mutants with smaller period effects showed mild starch excess phenotypes that were not so readily interpretable, but with small-effect mutants, the variability among experimental replicates also limited the interpretation of starch profile data. Highly revealing data were produced using ‘an unexpected early dusk’ treatment, which tested for instantaneous slowing of the starch degradation rate when plants experienced a long, 16h night for the first time. Altered temperature treaments (Section 1b) and light intensities (Sections 2.c.iii 5.f) allowed us to manipulate the plant’s starch content independently of photoperiod duration.
Significant results:
• No clock mutant with the possible exception of gi deviated much from the constant starch degradation rate throughout the night, even if they altered the average rate.
•
• Starch degradation rate could slow below the ‘dawn-prediction’ rate, when plants were sugar-replete.
•
• Starch degradation rate failed to accelerate above the ‘dawn-prediction’ rate, when the plants were starved for sugar.
•
• Most clock mutants and starch pathway mutant plants retained the remarkable, instantaneous slowing of starch degradation in response to the unexpected early dusk.
•
• Mutants of the starch-phosphorylating enzyme phosphoglucan water dikinase (PWD), in contrast, showed no change in starch degradation following the early dusk.
•
Several lines of evidence suggested that the clock controls starch degradation. These results (and model-guided analysis, see Section 2d) indicated that none of the clock components tested (by mutation) individually or directly provided the timing input to set starch degradation rate. Our working hypothesis was that this control was indirect and depended upon multiple clock components. Our results included two major discoveries. First, the clock-determined upper limit on starch degradation rate is robust; even starvation signals cannot override this limit. However, the potential redundancy of its circadian components limited our investigation of the timing mechanism starting from the clock. Second, the result in the pwd mutant strongly suggested that the cycle of phosphorylation/de-phosphorylation of the starch surface is a possible target for the signalling mechanisms that link the clock to starch degradation.
ii. The Tre6P signalling intermediate
A further discovery of TiMet concerned the signalling metabolite trehalose 6-phosphate (Tre6P), which we showed is an inhibitor of starch degradation. A rich data set was acquired by testing Tre6P levels in the samples of Section 1b, using transgenic plants with constitutive or inducible Tre6P production or reduced Tre6P level, independently manipulating sucrose levels by using transporter mutant plants that accumulate sucrose, and manipulating starch formation using mutants defective in the synthesis of normal starch granules.
Significant results:
• Tre6P tracks sucrose availability.
•
• Starch degradation in the night is inhibited by Tre6P.
•
• Tre6P acts to inhibit starch degradation at a point close to the start of the degradation pathway.
•
Thus in plants with abundant starch and high sucrose (e.g. sink-limited plants, in 18h photoperiods), increased Tre6P can reduce starch degradation rate below the clock-limited maximum.
• In the conditions where Tre6P limits starch degradation, the starch degradation rate is temperature-sensitive, whereas the clock-limited, maximal rate is temperature-insensitive (see Section 2.a).
•
• Moreover, Tre6P was shown to interact with the clock in further ways, including regulation of clock genes and of clock outputs involved in the FT photoperiod floral initiation pathway (see Section 5e).
•
These discoveries reveal how starch mobilisation may be adjusted according to demand for sucrose during the night. In a given situation, the clock sets a maximum rate of starch degradation that avoids starvation until dawn. The actual rate of starch degradation may be lower. If more starch is available than can be used for growth and maintenance, sucrose will increase resulting in an increase of Tre6P and inhibition of starch degradation. We have now developed the concept of a sucrose-Tre6P nexus to describe the relationship between these two molecules: they are not only interdependent signals, as each can also trigger signalling processes independently of the other.
iii. Detailed starch dynamics and complex light regimes
After observing the more detailed dynamics of the starch regulation under different photoperiods, the TiMet project deepened our analysis in three directions. We aimed to test, first, which particular aspects of the light:dark cycle entrained the circadian clock, and thus set the phase of subjective dawn, the target for the clock-limited starch degradation rate. When confronted with a “double-dawn”, in which light levels increased in two steps from dark to very low to high, the clock entrained to a time that lay between the two steps (see section 1). Second, we tested a mechanism to explain the more complex profile of starch accumulation observed in 18h photoperiods. Third, we tested which aspect of the light:dark cycle triggered that mechanism.
Significant results:
• The plant clock integrates low- and high-light signals in entrainment. This not only affected the entire clock system but also set the timing of starch degradation in the subsequent night.
•
• In long photoperiods, starch degradation decreases gradually during the first hours of the light period and increases gradually before dusk. This is also likely in natural light regimes (section 5).
•
• Initial results indicate that starch degradation can be initiated by falling light intensity. This response is modulated (“gated”) by the circadian clock, such it occurs only towards the start and end of the day.
•
This work represents a new level of detail in the understanding of starch regulation, which accounts not only for overall daily profiles but for dynamic patterns, hour by hour. Taken together, our results suggest that the joint, light and circadian regulation of starch degradation promotes a smooth transition from photosynthetic to stored carbon sources under natural light regimes. These mechanisms together prevent the transient starvation during the transitions between light and darkness, which we observe in laboratory “square-wave” light cycles.
d. Models of starch metabolism (WP2.2 2.4 3.3)
Several different classes of models of the diurnal and circadian regulation of starch metabolism were developed in TiMet, far exceeding the scope of any other results in this area. The key insight started immediately before TiMet and was validated by TiMet’s experimental data. A simplified phenomenological framework was then developed, which allowed us to explore general principles of the diurnal and circadian regulation of starch metabolism. Two more detailed models then explored potential molecular mechanisms, generating hypotheses for future experimental tests. Unexpectedly, this work also led to a new theoretical framework with rate laws to describe surface-active enzymatic reactions, such as those occurring on the starch granule (Section 5).
Significant results:
• The control of starch degradation in different photoperiods implements “arithmetic division”. The rate of degradation is set by dividing the amount of starch at dusk by the time remaining until predicted dawn. The sensor for starch amount is unknown; the timer derives from the circadian clock.
•
• No model of a single, direct connection from a clock component to starch degradation is consistent with the data, suggesting that the connection depends on multiple clock components and is mediated by an intermediate regulator.
•
• The intermediate regulator could be either an activator or an inhibitor of starch degradation.
•
• The intermediate regulator is reset shortly after dawn, but not immediately, in a light-dependent fashion that is likely gated by the circadian clock.
•
• The SnRK1 kinase and its β subunit AKINβ1 might mediate the effects of this regulator, controlling both synthesis and degradation of starch.
•
• A separate protein-kinase-dependent mechanism might form the link to circadian timing, and also link the system to Tre6P.
•
• Circadian control of the protein kinase signalling could coordinate carbon supply with clock-regulated control of growth, which affects carbon demand.
•
The suite of models parallels the development of TiMet’s understanding, starting from overall regulation of starch metabolism across photoperiods, and moving to an hour-by-hour analysis of system dynamics that depends upon integrating the central metabolic pathways with the control of growth (see section 4).
3. New understanding of isoprenoid metabolism
The diverse families of isoprenoid chemicals are produced either in the cytosol of plant cells or in plant plastids. Cytosolic isoprenoid products include the key hormones, cytokinins and brassinosteroids, as well as other sterols and enzyme cofactors, which are derived from the MVA synthesis pathway. Plastidial isoprenoids include chlorophylls, and include essential functions in photosynthesis and photoprotection. The plastidial isoprenoids are derived from the MEP metabolic pathway in the plastid. Precursor compounds are diverted from the MEP pathway into the various synthesis pathways for the particular isoprenoid compound families. In addition to the early steps of the MEP pathway, TiMet research focused on carotenoids and tocopherols, two families of plastidial isoprenoid metabolites with antioxidant properties.
a. Genes of the isoprenoid pathways
RNA levels of several isoprenoid pathway genes were known to be rhythmic prior to the project. As the pathways have many branches that synthesise compounds of distinct chemical families, so some genes are present in multiple copies. Synthesis of the key precursor geranylgeranyl diphosphate (GGPP) is encoded by 12 predicted GGPPS genes in the A. thaliana genome, for example: TiMet studied their function and the sub-cellular localisation of their proteins.
i. Isoprenoid pathways in transcriptome and proteome analysis (WP1.3; WP2.3)
TiMet’s detailed timeseries data (Section 1b) showed that mRNA transcripts for plastidial isoprenoid enzymes were more abundant in the morning than in the evening, especially for the enzymes that divert precursors into specific pathways (PSY for carotenoids, GGR for tocopherols) and two of the most important MEP enzymes (DXS and HDR). Our data were consistent with published results suggesting morning-specific, rhythmic regulation of these genes. Only two mRNAs from these pathways were evening-expressed: VDE catalyses a reverse reaction in terminal carotenoid synthesis; VTE2 is the first committed enzyme of tocopherol biosynthesis.
The global transcriptome datasets generated by TiMet allowed us to identify putative transcriptional and post-transcriptional regulatory links between the circadian clock and a wider set of target genes in the isoprenoid pathways. The analysis considered both RNA expression profiles and the regulatory sequences present in the upstream regions of genes encoding the MVA and MEP pathway enzymes. To test potential retrograde effects of the isoprenoid pathway metabolites upon the clock, plants mutated for MVA pathway genes (aact1-1, hmg1-2, hmg2-1, mk, mpdc1-1 and mpdc2) were also subjected to transcriptome analysis. The latter studies were extended to climate change treatments (see Section 5.f).
Significant results:
• 10 of the 12 GGPPS genes in Arabidopsis encode functional proteins. Their specific subcellular location and differential expression pattern suggest that the gene family is maintained due to sub-functionalization, such that different genes provide GGPP to specific tissues, developmental stages, or metabolic pathways.
•
• HUD and EE promoter sequences (linked to RNA expression during the day phase) are found upstream of both MVA and MEP pathway genes, potentially explaining morning-peaking expression of several MEP pathway genes.
•
• In contrast, the expression of MVA-pathway genes is inhibited by light.
•
• For MVA pathway enzymes encoded by multiple genes, in general only one of the genes has circadian and/or light specific cis-elements in the promoter region. Light and/or circadian regulation might thus be decoupled from other regulatory processes (e.g. expression in leaves vs. roots).
•
• The global proteomics data sets indicated that key enzymes of the MVA pathway (HMGR) and of GGPP biosynthesis (GGPPS) were more abundant in long photoperiods.
•
• RNA levels for clock genes LHY and CCA1 were each altered by one of the MVA pathway mutants (mdpc2 and hmgr2, respectively).
•
• RNA levels for the transcription factor HY5, which is known to bind CCA1 and ELF4 hence controlling circadian gene expression, are altered in hmgr1 and hmgr2 mutants.
•
ii. Mechanisms of transcriptional regulation (WP2.3 2.5)
To understand in more depth the control of synthesis of plastidial isoprenoids and their downstream products, we examined the roles of the light- and clock-regulated HY5 and PIF family transcription factors. RNA assays, reporter gene fusions and mutant plants lacking these TFs all contributed to the analysis. The HY5 and PIF TFs are also included in our models of clock function (e.g. section 4b).
Significant results:
• HY5 and PIF1/PIF4 impart antagonistic regulation to common gene targets involved in the production of carotenoids and chlorophylls, through direct binding of both TF classes to the same G-box promoter element.
•
• The Arabidopsis PSY gene is regulated by PIFs only in response to light signals in the shoot, and is neither PIF-regulated nor light-responsive in the root.
•
This dynamic transcriptional module acts with the circadian clock to adjust the level of rhythmic gene expression. Sudden changes in either light or temperature adjust the equilibrium of HY5 and PIF bound to target promoters, altering the transcriptional response. In this way the expression of genes involved in the biosynthesis of MEP-derived plastidial isoprenoids, such as chlorophylls and carotenoids, can be modulated by external signals.
b. “Anti-rhythmic” regulation of DXS protein levels (WP2.3 2.5)
A straightforward interpretation of the morning-peaking RNA levels of the genes in the carotenoid pathway would be that carotenoid synthesis is also most active in the morning. A physiological advantage was previously proposed for this regulation, namely increased protection against predictable light stress, promptly from the start of the light period. However, TiMet’s analysis of multiple tocopherol and carotenoid compounds showed no rhythmic regulation under constant light conditions, presenting a paradox: RNA levels were clearly and co-ordinately rhythmic but the levels of the metabolic products were not. Our work to resolve this paradox revealed a previously-undescribed mode of “anti-rhythmic” protein regulation, with evidence that this regulation is important for plant growth.
Significant results:
• DXS and HDR proteins were shown to be rapidly degraded, in isolated chloroplasts, suggesting that the rhythmic expression of their RNAs would lead to a morning peak in protein levels.
•
• DXS degradation required Clp proteases in the chloroplast, mediated by a J-protein chaperone that likely targets misfolded forms of DXS.
•
• DXS protein level in wild-type plants changed very little over the normal light:dark cycle.
•
• Surprisingly, DXS protein levels did oscillate in transgenic plants bearing a constitutively-expressed, 35S:DXS transgene. In these plants, the protein level peaked in the evening.
•
These results suggested that the morning peak of rhythmic DXS RNA levels is normally counteracted by greater accumulation of DXS protein in the evening. The rhythmic RNA and ‘anti-rhythmic’ protein regulation results in the constant protein levels observed in wild-type plants.
• The growth defect of dxs knock-out mutant plants cannot be rescued with a constitutively-expressed 35S:DXS transgene (the resulting plants grew slowly). 35S:HDR and 35S:PSY constructs could not rescue hdr and psy mutant growth at all, resulting in a seedling-lethal, albino phenotype like dxs.
•
This result is consistent with the notion that constant levels of these proteins are important for normal plant growth. The tentative conclusion is that the clock-regulated RNA expression is essential to avoid a detrimental oscillation in enzyme levels. Presumably, the constant levels of enzymes are crucial to maintain constant levels of the final isoprenoid products, which are important for plant growth.
c. Regulation of DXR protein by the clock (WP2.3 2.5)
No changes in transcript levels were found for the gene encoding DXR, another rate-determining enzyme of the MEP pathway. However, levels of DXR protein (but not RNA transcript) were increased in toc1-1 mutant plants, which have a 20h biological clock due to the defect in the clock component TOC1. This molecular effect made the plants resistant to the DXR inhibitor compound, fosmidomycin, and suggested that the circadian clock was involved in regulating DXR.
Significant results:
• Several other clock mutants and a clock-affecting transgenic line also increased resistance to fosmidomycin, indicating that the result was not specific to toc1.
•
• Growth in 20h light:dark cycles restored the toc1 mutant to fosmidomycin sensitivity, as expected if the mutant’s phenotype was due to the mutant clock’s abnormal phase in 24h cycles.
•
• DXR protein levels were constant across the day in wild-type plants, and similar in 12h and 16h photoperiods. In toc1 mutants, DXR levels were higher at dawn, especially in 16h photoperiods, suggesting that the clock influences DXR protein stability. The DXS protein was not affected.
•
• DXR protein levels were shown to be regulated by Cpn60 chaperonins in wild-type plants, distinct from the mechanism regulating DXS.
•
d. Models of isoprenoid metabolism (WP2.3 2.4 2.5)
TiMet first developed metabolic pathway models of all currently known genes for enzymes in isoprenoid metabolism (MVA, MEP and associated pathways), annotated with the relevant Enzyme Catalogue numbers and linked to experimental data on the cognate Arabidopsis genes. These results were presented online in the AtIPD resource (Section 5a).
TiMet developed the first model of the MEP pathway of plastidial isoprenoid synthesis, integrating several mechanisms that influence the daily rhythms of the pathway. First, stimulation of the pathway flux by light through the supply of the substrate glyceraldehyde 3-phosphate (GAP) from photosynthesis. Second, feedback regulation of the key enzyme 1-deoxy-D-xylulose 5-phosphate synthase (DXS) by MEP-derived products. Third, circadian regulation of three steps in the pathway (DXS, HDR and consumption of ultimate products by other pathways). Regulation in sink and source tissues was distinguished.
Significant results:
• Enzymatic steps of the MEP, MVA and downstream isoprenoid pathways were annotated with the relevant Enzyme Catalogue numbers, AGI gene identifiers, and relevant proteomics data.
•
• The dynamic MEP model suggests that circadian regulation of the MEP pathway only slightly stimulates flux when plants are exposed to light during the day.
•
• Changes in GAP substrate availability from photosynthesis in source tissues are predicted to alter the MEP pathway flux much more strongly between day and night, compared to clock regulation.
•
• In sink tissue, circadian control might have a greater effect, comparable to the extent of substrate-level control, because substrate availability is less dependent on light.
•
• Jointly, light-dependent substrate and clock-modulated enzyme activities ensure that sufficient MEP product are available during the light phase of a day, when their demand is highest.
•
4. New understanding of plant growth.
a. Day/night balance of growth (WP1.1 1.2 2.5 3.1 3.3)
Growth in developmental biology is often taken to mean an increase in size, which is easily measured by non-destructive imaging. In TiMet, we focussed on growth as the conversion of carbon into cellular components. Growth in this context means biomass. 90% of fresh biomass in plants is water, which can be affected by many processes other than the metabolism of interest to us. Therefore the most relevant measure for us is dry biomass, which can only be measured destructively. Moreover, the composition of biomass in terms of cell wall, protein, etc., might in principle vary. We developed methodologies based on whole plant carbon balances, flux analysis to different organs and biomass components, including detailed analysis of amino acid pools and protein synthesis, and on analyses of ribosome abundance and polysome loading, that provide quantitative information about growth rates and mechanisms in the daytime and night.
Significant results:
• the overall rate of growth in a 24 h cycle is dependent on the net availability of carbon.
•
• the diurnal timing of growth is highly flexible, such that total growth in darkness can be as great as growth during the day.
•
• growth at night is determined by the amount of starch at dusk, the duration of the night and the rate of maintenance respiration,
•
• protein synthesis at night accounts for a large part of the total carbon and energy that is available from the degradation of starch,
•
• clock-dependent timing of starch degradation leads to stable changes in the sucrose level across most of the night,
•
• changes in sucrose content correlate with polysome loading (a proxy for the rate of protein synthesis),
•
• the daytime or night growth rate correlates with the corresponding content of sucrose, Tre6P and total amino acids,
•
• slow growth at night can lead to a delay before rapid growth is established on the following day
•
• in extreme cases, the delay occurs with signs of protein breakdown by a starvation response.
•
• carbon allocation is highly flexible during the 24 h cycle, consistent both with estimates from component reactions and with direct measurements of the flux to different biomass components,
•
• carbon allocation to different organs varies over plant development,
•
• the clock orchestrates ongoing rhythms of biochemical growth even in constant conditions,
•
• altered circadian timing of growth is correlated with effects on carbon metabolism.
•
These results led to a new, systems-level perspective, in which the quality, depth and breadth of data are sufficient to analyse the levels of central carbon metabolites (and increasingly, protein synthesis) as emerging from the balance of multiple influences upon the fluxes of supply and demand. This perspective was elaborated first on a daily and now on an hourly timescale.
b. Connecting clock outputs and the Framework Model (FM); WP3.1 3.3.
To represent the complex interactions uncovered by our results, and to provide a tool to further future understanding, TiMet built the first multi-scale, mathematical model of whole-plant growth for Arabidopsis thaliana. One component was to connect the model of the clock gene circuit to its best-characterise molecular outputs, the photoperiodic flowering pathway (including FKF1 and FT, see sections 2.b.i and 5e) and the light-regulated elongation pathway (including the PIF factors, see section 3.a.ii). The second was an integrated Framework Model (FM) that links gene network dynamics to the organ and whole plant levels, thus providing a means to understand the combined effects of circadian control and environmental signalling on carbon partitioning into starch, starch degradation, resource allocation and ultimately plant form. Validating the model depended on the rich data from TiMet.
Significant results:
• the clock-output model predicted new connections and dynamic features in the FT and PIF pathways, which were validated by experimental data,
•
• the clock-output model was extended to control hundreds of photoperiod-regulated PIF target genes, giving it genome-wide scope for the first time.
•
• the FM was built by integrating four models from the literature, in a modular fashion.
•
• the FM made accurate biomass predictions in reference conditions, with minimal recalibration, matching new data from Edinburgh and from Golm.
•
• the model was extended to include stochastic development, and resolved unanswered questions on the phenotype of a developmental mutant.
•
• simulations of the FM show how our data invalidate the earlier ‘overflow’ hypothesis for allocation of ‘spare’ carbon to starch, and require a flexible, conservative strategy for starch storage under different photoperiods,
•
• the modular model structure allows individual sub-models to be replaced and updated easily,
•
• a version of the model was developed in a user-friendly, graphical modelling software, to facilitate dissemination and re-use; online tutorials were held for a live, international audience.
•
• integrating the clock-output model allowed us to simulate the pleiotropic effects of a long-period clock mutant, matching its effects upon starch degradation, elongation growth, flowering and biomass growth.
•
• other measured carbon reserve pools are affected by the mutant (section 1d) and are required, together with the starch defect, to account quantitatively for the decreased growth phenotype of this mutant.
TiMet has therefore demonstrated that it is possible to establish a quantitative link between a change in the dynamics of a gene regulatory network (in this case, the circadian clock) and a complex emergent phenotype like whole-organism biomass, for the first time in any large eukaryote. Our results indicate that existing sub-models produced in different laboratories can usefully be combined, despite the variation in plant growth observed among laboratories, so long as each laboratory’s experiments are appropriately documented. Thus many laboratories can contribute to a community-based effort to develop and extend such models.
c. Methods for spatial analysis (WP3.1 3.3)
In parallel with our focus on biomass growth, two new technical developments allowed us to measure elongation growth, contributing to our understanding of hourly timescales.
i. Root elongation
Root extension growth is an easily-measurable growth parameter in young Arabidopsis plants, because it occurs in only one dimension, in a single organ that elongates without other movement. Imaging studies revealed a circadian oscillation in root extension rate, with a trough at ZT8-10, which depends on the evening complex. The growth pattern was modified by carbon supply: low carbon leads to an almost immediate inhibition of growth. The importance of clock-regulation of starch degradation for growth was demonstrated by showing that premature exhaustion of starch in the short period lhy cca1 double mutant leads to an inhibition of root extension growth at the end of the night, which is recovered by adding exogenous sucrose. The root extension assay is therefore a time-resolved, physiological marker for carbon availability in the whole plant.
ii. Rosette expansion
Rosette expansion growth is a far more complex process, involving multiple organs that both expand and also move, in three dimensions. TiMet developed a novel method for spatially- and temporally-resolved analysis of rosette growth, using a newly-developed, commercial light-field camera. The method uses only one camera, greatly simplifying the process compared to competing methods. The equipment, with a significant development of image processing and analytical software, allows reconstruction of a 3D image of an Arabidopsis rosette at each timepoint, and hence corrects for leaf movement, allowing precise quantification of expansion growth in a freely moving plant. It is now being applied to test the range of samples and conditions of interest to TiMet.
5. New research methods and extensions to unexpected areas.
The creativity and reach of our partners allowed the project to return exciting results in unexpected areas relevant to the project.
a. Data infrastructure and analysis methods (WP1, WP2.4 2.5)
TiMet’s data management infrastructure reflected the diversity of large-scale experimental methods and modelling methods used in the project.
• A comprehensive and interactive Arabidopsis Isoprenoid Pathway Database (AtIPD) was constructed (http://www.atipd.ethz.ch/) which includes TiMet’s metabolic pathway models of the isoprenoid pathways, linked to the relevant Enzyme Catalogue numbers, data on the cognate Arabidopsis genes and subcellular localisation. The isoprenoid pathway models can be downloaded in several formats including for the Cell Designer modelling tool.
•
• A new relational database, TiMet-DB, was constructed to allow powerful queries across multiple, genome-scale data sets, including the TiMet transcriptome and proteome data. Analysis routines can be added to the database, organising both primary and secondary data.
•
• The BioDare online resource (www.biodare.ed.ac.uk) was extended with the most powerful and specific tools available to retrieve desired data from among hundreds of similar experiments, allowing sharing among collaborators and public dissemination. From two methods initially, BioDare’s functions were extended to 6 mathematical methods for online analysis of circadian rhythm data. New visualisation methods enhanced data display. RNA and metabolite timeseries data from several labs were stored in the repository.
•
• The PlaSMo model repository (www.plasmo.ed.ac.uk) was updated and hosts public dissemination of TiMet’s mathematical models, including the Framework Model.
•
b. Models for surface-active enzymes (WP2.4)
TiMet’s work on starch differs from most biochemistry in one important respect: the starch granule is a quasi-crystal. Enzymes that metabolize starch do not do so in solution, but on the granule surface. TiMet first considered how entropy affects the activities of enzymes that act on carbohydrate polymers like starch. Applying these ideas to the starch surface, however, prompted TiMEt researchers to develop a much more general, theoretical framework that describes surface-active enzymatic reactions. The new theory derives rate laws analogous to the classical Michaelis-Menten rate law for enzymes acting in solution. This development was not anticipated at the start of the project but it should be broadly applicable to many enzymes that function on solid surfaces, such as the surface of the starch granule.
c. Economic analysis (WP3.3)
Many results suggest that plants partition their resources in an optimal way, which resembles an economic optimisation of profit. This concept allowed a novel, conceptual approach to understand plant growth, which we expect to have broad relevance. TiMet considers the partitioning of growth between day and night, and of carbon between sugar for use and starch for storage. During the day, energy is freely available from photosynthesis, whereas at night, energy has to be remobilised from starch and other stores and is thus more expensive. However, the process of growth also requires machinery, which must be built using considerable resources. Thus it may be ‘worth’ growing during the night, even if energy costs are higher, rather than letting the valuable machinery lie idle. The analogy is with a capital-intensive factory, where it is advantageous to run a night-shift even if the wages during the night shift are higher than during the day. In preliminary studies, these economic principles were embodied in a mathematical model, in order to find the optimal balance of day- and night-time growth, depending on the energy invested in the growth machinery. Strikingly, the model’s predicted optimal strategies were the opposite of the experimental observations, indicating that other factors are important in the plant’s strategy. The optimality-based approach promises to guide future understanding of plant growth regulation, by revealing the factors that allow the model to match the data.
d. Translational Coincidence (WP2.4)
Data and insight from four TiMet partners contributed to discover an operating principle for plant gene expression, which we term “translational coincidence”. This molecular mechanism contributes to answer a fundamental question posed in our original application, namely how daily rhythms mediated by the circadian clock contribute to seasonal adjustment of plant metabolism under different photoperiods.
The concept was originally developed from TiMet observations of changes of protein abundance under different photoperiods and of different rates of translation in day and night (sections 1.c 4.a). The translational coincidence mechanism predicts a change in the levels of stable proteins in different photoperiods, if their translation rate in light is greater than in darkness, and if the transcript levels are rhythmic and controlled by a circadian clock. In Arabidopsis, the proteins encoded by rhythmic RNAs with peak levels in the evening were predicted to become more abundant under long photoperiods, whereas proteins from morning-peaking RNAs would become relatively less abundant. This prediction, along with several more specific, secondary predictions, was validated by hundreds of proteins in TiMet’s quantitative proteomics data and also by data from two independent projects. This insight may contribute to resolve a longstanding paradox: many rhythmic RNAs encode very stable proteins that do not change in abundance over the course of a day. Translational coincidence would allow these daily rhythms to control seasonal changes in protein abundance, where the time of the daily RNA peak determines the season of peak protein expression. Preliminary results suggest that this mechanism may operate generally, across all photosynthetic organisms, giving it far-reaching implications.
e. The Tre6P signal controls flowering as well as vegetative growth (WP2.2)
During the project, we discovered that the sugar signalling molecule Trehalose-6-Phosphate (Tre6P; see section 2.c.ii) interacts with another important clock output, namely the floral induction pathway. Plants engineered to increase T6P were known to be early flowering, whereas transformants with reduced Tre6P were late flowering. Thus the Tre6P signal might mediate the longstanding suggestion that sugar status can control flowering time, which previously had no known molecular mechanism. Moreover, it was unknown how this effect interacted with other known signalling pathways that regulate floral initiation in response to environmental signals like photoperiod and temperature. By analysing expression of the clock output genes GI and CO, and the floral mediator FT, in lines with altered Tre6P, we showed that Tre6P is not required for clock-dependent rise in GI and CO transcripts, but is required for the increase in FT transcript. Genetic analysis that tested the interdependence of the GI- and Tre6P-dependent regulation, led to the suggestion that Tre6P mediates a carbohydrate-dependent modulation of flowering, sensitising flowering to photoperiod when sucrose is high, and attenuating the response when sucrose is low. Thus the plant’s sugar status signalled via Tre6P interacts with circadian clock outputs to control the developmental transition to flowering, analogous to its effect on starch degradation to control vegetative growth.
f. Towards natural regimes: natural light, double-dawn and climate change scenarios (WP3.3)
Almost all previous experiments performed on the clock and on starch metabolism use a simple step light regime, which changes instantaneously from darkness to full light intensity and back. However, in natural light regimes the light intensity changes gradually. TiMet results led us to investigate how the clock is entrained by light in natural regimes and how this affects carbohydrate metabolism. Clock gene RNA levels, starch and key metabolite levels were tested in plants grown in regimes providing periods of high light (adequate for photosynthesis) and very low light (able to entrain the clock), in natural light regimes, and in regimes in which light levels fall abruptly at different times of day. Clock dynamics were modified in an unforeseen manner in natural light regimes, especially during the night, when a secondary feature of dusk and evening complex genes was greatly enhanced. The natural light also delayed the phase of the clock, and the timing of starch degradation. We propose that this delay will allow plants to maintain a sugar supply from starch in natural light environments where the light intensity rises gradually, such that ‘circadian dawn’ occurs before ‘photosynthetic dawn’.
Natural conditions are expected to change under predicted global climate scenarios, which involve increases of temperature and CO2 as well as water shortage in many areas of the globe. A large-scale, multifactorial experiment extended TiMet’s work, in collaboration with the European Science Foundation MOMEVIP project, to test the response of Arabidopsis wild type and isoprenoid mutants to increased temperature, CO2 and drought conditions. Detailed phenotypic, physiological and volatile data were acquired by the partners, while TiMet performed transcriptome analysis and ongoing proteome measurements. Among the genome-wide analysis, a striking observation was the loss of transcript fluctuations for three core clock genes between the end of the day (EOD) and the end of the night (EON) in the most severe, multiple-stress condition. These effects, however, disappeared when plants were re watered.
Potential Impact:
The TiMet project pioneered the dynamic, systems-level analysis of combined metabolic and gene-regulatory networks and their multiple links to growth. The project aimed to develop fundamental understanding of this important system, as a test case for using model-guided, mechanistic tools to analyse biological regulation across scales. This was only possible in the laboratory model species, Arabidopsis thaliana. Establishing the toolset and proof-of-principle in this species, however, will now justify and direct future work to apply this systems biology approach to other biological processes, and to other species including the best-known crops.
The partners maintained a close review of the emerging results from their work with a view to protecting the results, in line with the consortium agreement and as appropriate for a fundamental research project of this nature. No patent filings or other intellectual property protection were initiated by any partners from work on the project.
To engage and inform the stakeholders who will be impacted by our work in future, TiMet attended or organised four meetings with a broad range of stakeholders, in addition to the ongoing activities of individual partners in this and related research areas:
• Presentation to stakeholders at the FESPB and European Plant Science Organisation Plant Biology Congress, Freiburg, 2012, with over 1000 participants from 60 countries.
• Attendance at events of the EU COST action PlantEngine, 2013.
• Organisation of a half-day ‘research programming’ workshop in Barcelona in 2013 for the TiMet Stakeholder Platform, attended by 7 EU FP6 or FP7 projects in Plant Systems Biology, along with 7 companies and other organisations, in addition to all TiMet partners.
• Publication of a professionally-written “Success Story” for DG Research & Innovation, 2014.
• Presentations to the TiMet Stakeholder Platform attending a half-day workshop, organised before the TiMet final meeting, in Dusseldorf 2015.
All partners participated actively in dissemination events in many forums, communicating their results to the scientific community through conferences, seminar and publications.
Public dissemination was undertaken in events ranging from science fairs to public lectures, in addition to the public website www.timing-metabolism.eu and micro-blogging on the Twitter platform (@TiMet_project). The website was redesigned around the mid-term of the project to adopt a ‘weblog’ (aka blog) format, with short articles on the project’s work and related items of interest, organised in multiple themes, allowing frequent updates by all partners.
A detailed list of these activities is provided separately.
List of Websites:
http://www.timing-metabolism.eu
The ambition and scope of TiMet was unprecedented in the field of plant systems biology. The overall aim was to understand the growth rate of plants quantitatively, and in particular, how the circadian clock controlled key metabolic pathways to change plant growth rate over the day-night cycle and across the seasons. TiMet participants combined the methods and insights from multiple disciplines, from studies of enzyme activity in metabolism, to genome-scale molecular biology and plant physiology, and from statistical to mechanistic and multi-scale modelling approaches.
A test case for plant systems biology
Three, closely-linked biological systems were selected for the project. The interconnected genes of the 24-hour biological clock were already a paradigm for systems biology. The pathways of starch metabolism provide carbon storage that is essential to allow plant growth at night. Metabolic products of the isoprenoid pathway include chlorophyll and key plant hormones. The key achievement was to move from disjoint understanding of these separate systems in a static context, to dynamic, systems-level understanding linked quantitatively to the growth of the whole plant.
New results, methods, models and understanding
1. The circadian clock.
TiMet’s models provided a new framework to understand the dynamics of clock gene expression, integrating key clock genes and linking to multiple clock-output pathways for the first time in any organism. Transcriptome and proteome data showed how daily rhythms of molecular activity are affected by changing photoperiods, metabolic and clock mutants, and in climate change scenarios.
2. The starch metabolic pathway.
Molecular mechanisms that adjust the rates of starch synthesis and degradation under different photoperiods were discovered. Unexpectedly, starch degradation continues during the day. TiMet’s results explain the regulation of starch on an hourly timescale, and point to system-level feedback from growth, whereas previous understanding contrasted two, uniform states for day and night.
3. The isoprenoid metabolic pathway.
Rhythmic regulation of the isoprenoid pathway was characterised at the transcriptional and post-translational levels, identifying key regulators in the chloroplast, and integrated into the first mathematical model of this system.
4. Plant growth.
The project documented for the first time the remarkable flexibility in the partitioning of plant growth between day and night, revealed how energy and carbon utilisation are optimised, and quantified how global factors control growth rate. In our multi-scale Framework Model, a simulated clock mutant explained the mutant plant’s biomass, representing the first time that altered gene expression dynamics have been linked quantitatively to such a high-level property in any organism.
5. Extensions to unexpected areas.
TiMet’s analysis of plant growth was extended to natural conditions, to understand natural genetic variation within and across species, and formulated as a problem of micro-economic optimisation.
In addition to the many particular results, TiMet research validated three new and general operating principles for biological timing. Starch degradation implements “arithmetic division” to set the upper limit on the nightly starch degradation rate. The isoprenoid protein DXS revealed “anti-rhythmic” regulation, where transcriptional and post-translational rhythms in opposite phases combined to maintain a constant protein level. “Translational coincidence” uses daily rhythms in RNA level to control the abundance of hundreds of proteins, not within the day but across the seasons.
Project Context and Objectives:
Photosynthesis provides all of our food, and a significant proportion of our fuel and industrial raw materials. Demand for the products of photosynthesis will increase dramatically in the next few decades. At current rates of crop yield improvement, the FAO predicts that the world will face a major shortfall (ca. 30%) in the supply of primary foodstuffs relative to demand by mid-century. Optimising photosynthesis is a promising route to increase crop yield potential and a current target for research ranging from ecology to synthetic biology. However, the Earth’s rotation predictably removes sunlight – and hence photosynthesis – for a significant part of each day, and each plant’s lifespan. Plants must orchestrate the accumulation and utilisation of photosynthetic products over the daily cycle to avoid periods of starvation, and thus optimise growth rates. The many other plant processes that are directly or indirectly affected by the day/night cycle must be regulated in parallel, to avoid conflicting requirements and to optimise plant productivity. The TiMet project aimed to understand how plants achieve this daily optimisation.
Project context
Before the TiMet project, we and colleagues had shown:
(i) that plants possess a circadian clock, comprising a circuit of regulatory genes that provide predictive, temporal regulation of many plant processes over the day/night cycle,
(ii) that accumulation of starch provides a store of the sugar derived from photosynthesis that can be used at night, providing both cellular energy and a substrate for metabolic pathways such as isoprenoid biosynthesis that are important for growth.
(iii) that these systems are intimately interlinked. The circadian clock controls the rate of starch utilisation at night, so that this store lasts for the entire night, both in long, winter nights and in short, summer nights.
The loss of any one of these systems was known to result in serious decrease in plant productivity, so these biological processes were central to agricultural productivity. Importantly, the concentration of understanding, data, tools and resources available in the laboratory model plant Arabidopsis thaliana (and in this species only) made detailed studies of these processes experimentally tractable. The combined system was therefore a test case, in which to elucidate the coordination of responses of multi-level biological networks to incoming environmental signals. The was a strong case that starch metabolism might be subject to particular regulation that was not representative of other pathways, hence we included the isoprenoid pathway, which produces critical components for growth, from key cofactors such as chlorophyll, to antioxidants and plant hormones. We anticipated that these studies would require mathematical modelling to understand the complex, dynamic responses in gene regulation, metabolism and growth, and to link them across different time and length scales. However, the dynamics of a gene regulatory network had never been linked mechanistically, via altered metabolism, up to the level of whole-organism biomass, in any multicellular species.
At the start of the project, our understanding was disjointed and at disparate temporal resolution. The “chronobiology” and “metabolism” research fields were traditionally rather separate but now needed to be combined, in order to discover the mechanisms by which the clock exercises control over metabolism. Biochemical studies of detailed metabolic processes were also traditionally separated, both among areas of metabolism and from understanding of macroscopic growth. They too needed to be linked, in order to understand in depth how alterations of metabolism inhibited growth, in quantitative rather than qualitative terms.
These limitations were reflected both in experimental data and in the mathematical models available to formally represent our understanding of these systems, from (i) the operation of the circadian clock and its environmental light input, to (ii) the control of downstream clock outputs, to (iii) their joint effects on starch metabolism and in other downstream central metabolic pathways like isoprenoid metabolism, to (iv) the impact on whole plant growth. The dynamics of the circadian clock were studied at hourly resolution and were a paradigm for genetic responses to environmental signals. The daily, rhythmic regulation and negative feedback among the clock genes were also a paradigm for modelling of gene networks. The transcriptional regulators expressed at dawn LHY and CCA1 were well known, along with a family of PRR genes including TOC1, which regulated LHY and CCA1 by an unknown mechanism (they are now known to be direct transcriptional regulators). Additional components such as GI were known to link the clock with the pathway that accelerated flowering under long photoperiods. However, some of the most important clock components known could not be integrated into the models. In particular, mutants of ELF3, ELF4 and LUX genes had very severe effects up to abolishing circadian timing, but their biochemical functions were unknown (they are now known to form a repressor termed the Evening Complex). Models of the clock gene network had not been linked to any clock output, for any organism.
Starch metabolism in plants was studied at daily resolution, discriminating between two regimes. In the day, starch was thought to be synthesised but not degraded; in the night, starch synthesis ended and the degradation pathway mobilised sugars for use in the cell. Models of starch metabolism in other plant species had either been linked to photosynthetic inputs, with detailed biochemical mechanisms but without outputs to growth, or had been linked to growth but in a phenomenological context, without biochemical mechanisms. No dynamic models existed for isoprenoid metabolism. The only growth models with molecular detail focussed on developmental processes leading to shape alone, without any biomass, or accounted for biomass without molecular mechanisms. Within the modelling communities, research groups relied routinely on a subset of methods, rather than combining the multiple methods required at different stages of a systems biology investigation.
The TiMet project therefore assembled world leaders in experimental and theoretical plant systems biology, to understand the regulatory interactions between the clock gene circuit and plant metabolism, and their emergent effects on whole-plant growth. Many of the underpinning methods and resources for our studies of the Arabidopsis clock, starch and isoprenoid pathways were originally created by the TiMet partners. TiMet combined their exceptional and often world-leading, experimental facilities in order to address the problem at multiple levels of biological organisation, both in detailed assays and transcriptome- or proteome-wide. They were applied to analyse the responses to day-length and light-quality regimes that modify and disturb clock function in wild-type plants and in a large collection of mutant lines deficient in the clock, starch and isoprenoid metabolism. The data were organised in new infrastructure, and informed innovative modelling platforms that will permit functional analyses, generated further insights and testable hypotheses, which were then verified by further experiments.
Project objectives
Our objectives were to:
1. Establish a single experimental system to study clock, starch and isoprenoid subsystems, with mechanistic, mathematical models of each component, parameterised for this system.
2. Use high throughput, cutting-edge ‘omics platforms to acquire network-wide and targeted data on the network’s dynamic responses over the day/night cycle, with appropriate data handling to support data sharing and integration and binding to models.
3. Manipulate the network to discover the mechanistic connections from the clock to starch metabolism and isoprenoid synthesis, as well as inputs from these metabolic pathways that control the clock.
4. Include the new connections in the mathematical models of the component pathways.
5. Analyse both models and data in order to understand and ultimately predict the effect of each connection on the emergent properties of the intact system, including the carbon and energy available for long-term plant growth.
6. Test this understanding through dynamic manipulation of particular subsystems, monitoring outcomes at both the cellular and whole-plant levels.
7. Validate and refine the integrated model through iterative rounds of experimentation. Predict experimental strategies to further dissect the molecular network and to engineer/select desirable whole-plant behaviours, corresponding to crop traits.
The project aimed to develop fundamental understanding of this important system, as a test case for using model-guided, mechanistic tools to analyse biological regulation across scales. This was only possible in Arabidopsis. Establishing the toolset and proof-of-principle in this species, however, will now justify and direct future work to apply this systems biology approach to other biological processes, and to other species including the best-known crops.
Project Results:
The TiMet project achieved systems-level understanding of the links between circadian timing, metabolism and plant growth, crossing scales from intracellular pathways to the whole organism, and at a new level of temporal resolution. Over the 5-year lifespan of the project, the consortium has reached 37 Milestones and provided 33 Deliverables to the Commission. The consortium has published 52 research articles directly supported by TiMet, with 2 further papers in press, 7 book chapters or sections and 1 paper in conference proceedings. Many of these studies have appeared in the leading, international, peer-reviewed journals including Science, PNAS, Molecular Systems Biology, eLife, Plant Cell and influential review journals such as Annual Reviews of Plant Biology. In addition, 17 further publications are in preparation or under review for similar journals.
TiMet was a highly-integrated project, by design. Separating the results for presentation in any linear narrative necessarily results in multiple discontinuities. We therefore integrate the results from multiple work-packages into thematic sections below, in order to minimise these discontinuities, and note the relevant workpackages in the section headings. The close links that remain between the sections are indicated in the text. Moreover, new research methods were required in several cases to achieve the project’s ambitious aims. The development of these methods is usually described in the relevant results sections below or, for data infrastructures and one modelling method, in the final section.
1. New understanding of the biological clock and its pervasive effects.
TiMet aimed to develop a mathematical model of the clock genes, as they operated in the mature rosette leaves used in studies of starch metabolism rather than in the young seedlings typically used in studies of the circadian clock.
a. The circuit of genes that creates 24-hour biological rhythms (WP1, 2.4)
The existing, partial “P2010” model was first updated to represent newly-published findings. The key clock components ELF3, ELF4 and LUX were added, forming the “P2011” model, which also predicted a change in the function of the TOC1 protein. The prediction was verified experimentally by an independent group and by collaborators, who also showed that TOC1 controlled a greater than anticipated number of the other clock genes. An updated “P2012” model was then developed to include these data, explaining for the first time the clock gene expression pattern of the toc1 mutant. A stochastic version of the P2010 model showed how noise due to biological variability could affect the rhythmic gene expression.
To complement the subjective, expert judgement involved in building these models, statistical methods were also developed to learn the circuit connections that link the clock genes. This was a challenge for machine learning. Progress was aided by using simulated data from the stochastic P2010 model as a test case, because the true circuit underlying those data was completely known. Revised connections have now been suggested by these results.
Significant results:
• The P2011 model correctly predicted that TOC1 is a repressor, not an activator, of gene expression.
•
• Representing ELF3, ELF4 and LUX in P2011 changed our view of the clock’s operation.
•
• TiMet’s Hierarchical Bayesian Regression (HBR) learning method outperformed all competitors.
•
• An ensemble method, combining HBR and 14 others methods, outperformed all single methods.
•
b. Effects of environmental and genetic variation (WP1, 2.5)
The clock genes’ RNA profiles were measured in soil-grown plants of wild type (Col0 and Ws2) and circadian clock mutants (gi, toc1, prr7 prr9, lhy cca1 and elf3), and compared to data of collaborating researchers from seedlings that were grown with or without exogenous sucrose. Plants with inhibited or enhanced isoprenoid metabolism were also tested. The RNA profiles of the wild-type clock genes varied very little in these different conditions. The detailed behaviours of some clock mutants, however, suggested new connections among the clock genes. The absolute levels of clock gene RNA were measured for the first time. The wide range of expression levels was surprising, considering that all the clock genes are believed to function together in the same circuit.
Measuring the clock RNA profiles in plants grown with 4, 6, 8, 12 or 18h of light per day revealed how the change in day length progressively affected the clock’s timing. The “double-dawn” light conditions that were used in later studies (see Section 2.c.iii) were studied in greater detail, using our luciferase (LUC) reporter imaging facilities. These in vivo bioluminescence assays offer high temporal resolution and avoid the biological variation due to destructive sampling of different plants at each time-point for RNA. As little as 1µmol/m2/sec of light fluence rate at dawn was sufficient to set the phase of the rhythmic reporters. The LUC imaging assays also allowed us to test whether biological timing was uniform across the leaf. Leaves grown without environmental cycles for up to three weeks showed waves of rhythmic clock gene expression crossing the leaf, forming various spatial patterns. However, light-dark cycles fully synchronized all parts of the leaf, suggesting that the spatial patterns rarely occur in nature.
Arabidopsis Col-0 and two further accessions were also tested under five different temperature regimes, designed to investigate the impact of night temperature on metabolism and growth (see Section 2.a).
Significant results:
• Clock gene expression profiles were tested under the most diverse set of conditions ever reported.
•
• The first absolute quantification showed just ~50 PRR9 RNA copies/cell at the peak, the lowest level of the clock genes; the highest was LHY RNA, at over 1500 copies/cell.
•
• The clock model need not be specially adapted for rosette leaves, because clock gene expression patterns were very similar in wild-type tissue under all conditions, including with added sucrose.
•
• Almost all the progressive changes in phase under different photoperiods were matched by the existing clock models, indicating that these models are sufficient to analyse our metabolic data.
•
• The plant clock responded both to low intensity, photosynthetically-inactive light and to high-intensity light: their joint effects located subjective dawn to an intermediate time.
•
• Under constant light, different regions of a single leaf can differ in clock phase by up to 17h.
•
• Spatial patterns of clock phase indicate that clocks in neighboring regions are weakly coupled.
•
c. Testing for widespread, rhythmic regulation of mRNA and protein levels (WP1, 2.3 2.5)
The impact of the circadian clock on the broadest possible set of Arabidopsis mRNAs and proteins (the transcriptome and proteome) were analyzed in the wild type and clock mutants described above, and across multiple photoperiods, at dusk (end-of-day; EOD) and dawn (end-of-night; EON) time points. The disruption of core circadian clock components has a major impact on the transcriptome Changes in RNA transcript levels are not always reflected at the protein level. Thus, mutation of the clock genes does not globally affect the proteome but rather selectively influences specific protein targets.
Significant results:
• Altering the photoperiod from 6h to 18h has a profound effect: more mRNAs change in level at dawn in this comparison than change between dawn and dusk in the standard, 12h photoperiod.
•
• Fewer of the detectable proteins change expression between dawn and dusk, as expected, but the number changing in our results is greater than published estimates.
•
• Genes associated with the starvation (carbohydrate) response and encoding transcription factors (TFs) and protein kinases changed RNA level the most in the clock mutants. At the protein level, most changes were observed for carbohydrate response, phenylpropanoid, TFs and isoprenoid related genes.
•
• Proteins with a higher abundance in longer photoperiods included enzymes involved in the biosynthesis of glucosinolates, auxin and jasmonate hormones, and the starch degradation pathway. Proteins in the RNA processing pathway, lignin biosynthesis as well as in plastidial ribosomal proteins decreased in abundance.
•
d. Effects of daily timing on central metabolites (WP1, 2.1 2.5)
To test the effects of the clock and the altered conditions upon plant metabolism, photosynthesis, respiration, starch and key soluble metabolites were analyzed in the complete timeseries from wild-type plants, compared to clock mutants, altered temperature regimes and altered photoperiods, using the samples tested for the clock gene mRNAs (see 1.b above). These results provided the information to interpret starch metabolism in the context both of the carbon supply from photosynthesis and of its utilization in respiration and growth. Complementary metabolic labelling methods were also used to investigate the fluxes of carbon to protein synthesis, the other chemical constituents of biomass and the distribution among organs in whole plants (see part 4, below).
Significant results:
• Plants in photoperiods up to 12h were sugar-limited (“source limited”); plants in 18h photoperiod had excess sugar and were instead limited by the “sinks” that consume the sugar during growth.
•
• A previously undiscovered, circadian rhythm was detected in the level of a key sugar-phosphate.
•
• The lhy cca1 and prr7 prr9 mutants had the widest, and often reciprocal, impact on metabolite levels including sugar, amino and organic acid levels.
•
• Expert analysis of the entire data set suggested that distinct outputs from the clock affect starch degradation (controlled by multiple clock components; see part 2 below), organic acid and amino acid metabolism (controlled by morning clock components; see part 4), and reducing sugars (controlled by the evening complex genes).
•
• The machine learning ensemble method, including possible delays between RNA levels and the effect on metabolites, independently suggested results with a striking overlap: three clock genes connected to components of the starch synthesis pathway; one of the morning clock components was linked to levels of organic and amino acids; the suggested link from the evening complex was, however, to the sugar-phosphate rather than the unmodified sugar.
•
The striking results on starch metabolism are described in greater depth in section 2, with detailed analysis of the isoprenoid pathway in section 3, and the links from the clock and metabolism to growth in section 4, below.
2. New understanding of starch metabolism
Starch metabolism is critical for growth in Arabidopsis and many other plant species, because starch provides the major source of sugar to maintain cell function when darkness prevents C3 photosynthesis.
a. Effects of clock mutations, photoperiod and temperature (WP1.1 1.2)
The results from the clock mutants, altered photoperiod and temperature regimes (section 1b) on the timing of starch synthesis and degradation extended and reinforced the earlier results that prompted our studies of the starch pathway. Starch is synthesised at a higher rate in short photoperiod cycles, and more of the available photo-assimilate is partitioned into starch, compared to longer photoperiods. Starch degradation proceeds at a nearly constant rate from dusk to dawn, controlled to consume all but 10% (often as little as 5%) of the starch at dusk. The high-quality TiMet data provided the metabolic and growth context to understand the response of starch metabolism to many different manipulations.
Significant results:
• The clock-dependent timing of starch degradation led to stable but photoperiod-dependent sucrose levels across most of the night.
•
• Starch synthesis begins rapidly in the day, except in the 18h photoperiod, when there is a plateau for a few hours after dawn before starch levels start to rise.
•
• In 18h photoperiods, another plateau in starch synthesis occurs before dusk, when starch levels cease rising at the end of the day.
•
• Starch synthesis rates increased reliably under short photoperiods.
•
• Starch degradation rate was affected by photoperiod and clock mutants.
•
• Surprisingly, the starch degradation rate was hardly affected by night-time temperature.
•
• In sink-limited plants, however, starch degradation rate became temperature-sensitive.
•
The first conclusion was that Arabidopsis has a conservative strategy, storing more carbon as starch during short photoperiods in order to eke out these reserves during the following, long night. However, TiMet’s series of experiments (below) revealed considerable flexibility in some parts of this programme, and identified several of the mechanisms involved.
b. Control of starch synthesis (WP2.2)
i. Potential links from the clock
Both starch synthesis and degradation rates were predictably and reciprocally affected by photoperiod. Data on the circadian clock mutants tested the role of the clock in this photoperiod-dependent response.
Significant results:
• Most of the clock mutants that altered starch degradation rates (see below) did not affect the increase in starch synthesis rate in short photoperiods. Their pleiotropic effects, including starch excess phenotypes, might complicate these effects.
•
• In the gi and fkf1 mutants, however, starch synthesis did not respond to photoperiod. GI and FKF1 proteins form a complex essential for the clock-controlled photoperiod response of flowering time (see Section 4b).
•
The response of starch synthesis to changing day length is probably not directly controlled by the circadian clock components. The clock may still be indirectly involved, perhaps through a process involving GI and FKF1, which appear to form a novel link between the clock and primary metabolism. Both of these genes are strongly clock-regulated, with peak expression in the mid to late day. As the origin of the signalling mechanism from the clock was recalcitrant, the project focussed on identifying the target enzymes in the synthesis pathway that mediate the response.
ii. Target enzymes in starch synthesis
We hypothesised that photoperiod control of starch synthesis might be mediated by modulation of the activity of ADPglucose pyrophosphorylase (AGPase), an enzyme with several potentially important regulatory properties. Substituting altered forms of AGPase confirmed this hypothesis.
Significant results:
• The photoperiod response in starch synthesis was absent from plants expressing an unregulated, bacterial form of AGPase, though their starch profiles were normal in 12h photoperiods.
•
• The general pattern of starch accumulation in plants expressing a mutated, redox-insensitive, constitutively-active plant AGPase was similar to that of wild type-plants.
•
• Adjustment of the rate of starch synthesis to changing day lengths is not brought about by modulation of subunit composition, as mutants lacking individual subunits could still adjust starch synthesis to changing day lengths, and the immune-precipitated enzyme composition was the same in both photoperiods.
•
The regulatory properties of the plant AGPase are therefore essential for the response of starch synthesis to photoperiod but the redox modulation of AGPase activity was not necessary. The remaining well-understood regulatory property of AGPase is its allosteric activation and inhibition by 3-phosphoglycerate (3PGA) and inorganic phosphate (Pi) respectively. This property is believed to be important for the adjustment of the rate of starch synthesis in response to imbalances between sucrose synthesis/demand and carbon assimilation within a single day. TiMet’s results extend the role of allosteric regulation to the photoperiod-dependent increase in synthesis rate, which occurs without a major imbalance in synthesis and demand for sugar. This adjustment might reflect altered daily patterns of growth and hence demand for photosynthate in different day lengths, rather than a direct influence of the clock on the synthesis pathway itself, and is therefore discussed further in Section 4. This conclusion contrasts with more direct clock regulation of starch degradation.
c. Adjustment of starch degradation (WP2.1 2.2 2.5)
i. Potential links from the clock to target enzymes in the pathway
TiMet extended earlier work on the clock mutants. The strong, short- or long-period mutants clearly degraded their starch reserves to expire too early or late, respectively, relative to the actual time of dawn. The mutants with smaller period effects showed mild starch excess phenotypes that were not so readily interpretable, but with small-effect mutants, the variability among experimental replicates also limited the interpretation of starch profile data. Highly revealing data were produced using ‘an unexpected early dusk’ treatment, which tested for instantaneous slowing of the starch degradation rate when plants experienced a long, 16h night for the first time. Altered temperature treaments (Section 1b) and light intensities (Sections 2.c.iii 5.f) allowed us to manipulate the plant’s starch content independently of photoperiod duration.
Significant results:
• No clock mutant with the possible exception of gi deviated much from the constant starch degradation rate throughout the night, even if they altered the average rate.
•
• Starch degradation rate could slow below the ‘dawn-prediction’ rate, when plants were sugar-replete.
•
• Starch degradation rate failed to accelerate above the ‘dawn-prediction’ rate, when the plants were starved for sugar.
•
• Most clock mutants and starch pathway mutant plants retained the remarkable, instantaneous slowing of starch degradation in response to the unexpected early dusk.
•
• Mutants of the starch-phosphorylating enzyme phosphoglucan water dikinase (PWD), in contrast, showed no change in starch degradation following the early dusk.
•
Several lines of evidence suggested that the clock controls starch degradation. These results (and model-guided analysis, see Section 2d) indicated that none of the clock components tested (by mutation) individually or directly provided the timing input to set starch degradation rate. Our working hypothesis was that this control was indirect and depended upon multiple clock components. Our results included two major discoveries. First, the clock-determined upper limit on starch degradation rate is robust; even starvation signals cannot override this limit. However, the potential redundancy of its circadian components limited our investigation of the timing mechanism starting from the clock. Second, the result in the pwd mutant strongly suggested that the cycle of phosphorylation/de-phosphorylation of the starch surface is a possible target for the signalling mechanisms that link the clock to starch degradation.
ii. The Tre6P signalling intermediate
A further discovery of TiMet concerned the signalling metabolite trehalose 6-phosphate (Tre6P), which we showed is an inhibitor of starch degradation. A rich data set was acquired by testing Tre6P levels in the samples of Section 1b, using transgenic plants with constitutive or inducible Tre6P production or reduced Tre6P level, independently manipulating sucrose levels by using transporter mutant plants that accumulate sucrose, and manipulating starch formation using mutants defective in the synthesis of normal starch granules.
Significant results:
• Tre6P tracks sucrose availability.
•
• Starch degradation in the night is inhibited by Tre6P.
•
• Tre6P acts to inhibit starch degradation at a point close to the start of the degradation pathway.
•
Thus in plants with abundant starch and high sucrose (e.g. sink-limited plants, in 18h photoperiods), increased Tre6P can reduce starch degradation rate below the clock-limited maximum.
• In the conditions where Tre6P limits starch degradation, the starch degradation rate is temperature-sensitive, whereas the clock-limited, maximal rate is temperature-insensitive (see Section 2.a).
•
• Moreover, Tre6P was shown to interact with the clock in further ways, including regulation of clock genes and of clock outputs involved in the FT photoperiod floral initiation pathway (see Section 5e).
•
These discoveries reveal how starch mobilisation may be adjusted according to demand for sucrose during the night. In a given situation, the clock sets a maximum rate of starch degradation that avoids starvation until dawn. The actual rate of starch degradation may be lower. If more starch is available than can be used for growth and maintenance, sucrose will increase resulting in an increase of Tre6P and inhibition of starch degradation. We have now developed the concept of a sucrose-Tre6P nexus to describe the relationship between these two molecules: they are not only interdependent signals, as each can also trigger signalling processes independently of the other.
iii. Detailed starch dynamics and complex light regimes
After observing the more detailed dynamics of the starch regulation under different photoperiods, the TiMet project deepened our analysis in three directions. We aimed to test, first, which particular aspects of the light:dark cycle entrained the circadian clock, and thus set the phase of subjective dawn, the target for the clock-limited starch degradation rate. When confronted with a “double-dawn”, in which light levels increased in two steps from dark to very low to high, the clock entrained to a time that lay between the two steps (see section 1). Second, we tested a mechanism to explain the more complex profile of starch accumulation observed in 18h photoperiods. Third, we tested which aspect of the light:dark cycle triggered that mechanism.
Significant results:
• The plant clock integrates low- and high-light signals in entrainment. This not only affected the entire clock system but also set the timing of starch degradation in the subsequent night.
•
• In long photoperiods, starch degradation decreases gradually during the first hours of the light period and increases gradually before dusk. This is also likely in natural light regimes (section 5).
•
• Initial results indicate that starch degradation can be initiated by falling light intensity. This response is modulated (“gated”) by the circadian clock, such it occurs only towards the start and end of the day.
•
This work represents a new level of detail in the understanding of starch regulation, which accounts not only for overall daily profiles but for dynamic patterns, hour by hour. Taken together, our results suggest that the joint, light and circadian regulation of starch degradation promotes a smooth transition from photosynthetic to stored carbon sources under natural light regimes. These mechanisms together prevent the transient starvation during the transitions between light and darkness, which we observe in laboratory “square-wave” light cycles.
d. Models of starch metabolism (WP2.2 2.4 3.3)
Several different classes of models of the diurnal and circadian regulation of starch metabolism were developed in TiMet, far exceeding the scope of any other results in this area. The key insight started immediately before TiMet and was validated by TiMet’s experimental data. A simplified phenomenological framework was then developed, which allowed us to explore general principles of the diurnal and circadian regulation of starch metabolism. Two more detailed models then explored potential molecular mechanisms, generating hypotheses for future experimental tests. Unexpectedly, this work also led to a new theoretical framework with rate laws to describe surface-active enzymatic reactions, such as those occurring on the starch granule (Section 5).
Significant results:
• The control of starch degradation in different photoperiods implements “arithmetic division”. The rate of degradation is set by dividing the amount of starch at dusk by the time remaining until predicted dawn. The sensor for starch amount is unknown; the timer derives from the circadian clock.
•
• No model of a single, direct connection from a clock component to starch degradation is consistent with the data, suggesting that the connection depends on multiple clock components and is mediated by an intermediate regulator.
•
• The intermediate regulator could be either an activator or an inhibitor of starch degradation.
•
• The intermediate regulator is reset shortly after dawn, but not immediately, in a light-dependent fashion that is likely gated by the circadian clock.
•
• The SnRK1 kinase and its β subunit AKINβ1 might mediate the effects of this regulator, controlling both synthesis and degradation of starch.
•
• A separate protein-kinase-dependent mechanism might form the link to circadian timing, and also link the system to Tre6P.
•
• Circadian control of the protein kinase signalling could coordinate carbon supply with clock-regulated control of growth, which affects carbon demand.
•
The suite of models parallels the development of TiMet’s understanding, starting from overall regulation of starch metabolism across photoperiods, and moving to an hour-by-hour analysis of system dynamics that depends upon integrating the central metabolic pathways with the control of growth (see section 4).
3. New understanding of isoprenoid metabolism
The diverse families of isoprenoid chemicals are produced either in the cytosol of plant cells or in plant plastids. Cytosolic isoprenoid products include the key hormones, cytokinins and brassinosteroids, as well as other sterols and enzyme cofactors, which are derived from the MVA synthesis pathway. Plastidial isoprenoids include chlorophylls, and include essential functions in photosynthesis and photoprotection. The plastidial isoprenoids are derived from the MEP metabolic pathway in the plastid. Precursor compounds are diverted from the MEP pathway into the various synthesis pathways for the particular isoprenoid compound families. In addition to the early steps of the MEP pathway, TiMet research focused on carotenoids and tocopherols, two families of plastidial isoprenoid metabolites with antioxidant properties.
a. Genes of the isoprenoid pathways
RNA levels of several isoprenoid pathway genes were known to be rhythmic prior to the project. As the pathways have many branches that synthesise compounds of distinct chemical families, so some genes are present in multiple copies. Synthesis of the key precursor geranylgeranyl diphosphate (GGPP) is encoded by 12 predicted GGPPS genes in the A. thaliana genome, for example: TiMet studied their function and the sub-cellular localisation of their proteins.
i. Isoprenoid pathways in transcriptome and proteome analysis (WP1.3; WP2.3)
TiMet’s detailed timeseries data (Section 1b) showed that mRNA transcripts for plastidial isoprenoid enzymes were more abundant in the morning than in the evening, especially for the enzymes that divert precursors into specific pathways (PSY for carotenoids, GGR for tocopherols) and two of the most important MEP enzymes (DXS and HDR). Our data were consistent with published results suggesting morning-specific, rhythmic regulation of these genes. Only two mRNAs from these pathways were evening-expressed: VDE catalyses a reverse reaction in terminal carotenoid synthesis; VTE2 is the first committed enzyme of tocopherol biosynthesis.
The global transcriptome datasets generated by TiMet allowed us to identify putative transcriptional and post-transcriptional regulatory links between the circadian clock and a wider set of target genes in the isoprenoid pathways. The analysis considered both RNA expression profiles and the regulatory sequences present in the upstream regions of genes encoding the MVA and MEP pathway enzymes. To test potential retrograde effects of the isoprenoid pathway metabolites upon the clock, plants mutated for MVA pathway genes (aact1-1, hmg1-2, hmg2-1, mk, mpdc1-1 and mpdc2) were also subjected to transcriptome analysis. The latter studies were extended to climate change treatments (see Section 5.f).
Significant results:
• 10 of the 12 GGPPS genes in Arabidopsis encode functional proteins. Their specific subcellular location and differential expression pattern suggest that the gene family is maintained due to sub-functionalization, such that different genes provide GGPP to specific tissues, developmental stages, or metabolic pathways.
•
• HUD and EE promoter sequences (linked to RNA expression during the day phase) are found upstream of both MVA and MEP pathway genes, potentially explaining morning-peaking expression of several MEP pathway genes.
•
• In contrast, the expression of MVA-pathway genes is inhibited by light.
•
• For MVA pathway enzymes encoded by multiple genes, in general only one of the genes has circadian and/or light specific cis-elements in the promoter region. Light and/or circadian regulation might thus be decoupled from other regulatory processes (e.g. expression in leaves vs. roots).
•
• The global proteomics data sets indicated that key enzymes of the MVA pathway (HMGR) and of GGPP biosynthesis (GGPPS) were more abundant in long photoperiods.
•
• RNA levels for clock genes LHY and CCA1 were each altered by one of the MVA pathway mutants (mdpc2 and hmgr2, respectively).
•
• RNA levels for the transcription factor HY5, which is known to bind CCA1 and ELF4 hence controlling circadian gene expression, are altered in hmgr1 and hmgr2 mutants.
•
ii. Mechanisms of transcriptional regulation (WP2.3 2.5)
To understand in more depth the control of synthesis of plastidial isoprenoids and their downstream products, we examined the roles of the light- and clock-regulated HY5 and PIF family transcription factors. RNA assays, reporter gene fusions and mutant plants lacking these TFs all contributed to the analysis. The HY5 and PIF TFs are also included in our models of clock function (e.g. section 4b).
Significant results:
• HY5 and PIF1/PIF4 impart antagonistic regulation to common gene targets involved in the production of carotenoids and chlorophylls, through direct binding of both TF classes to the same G-box promoter element.
•
• The Arabidopsis PSY gene is regulated by PIFs only in response to light signals in the shoot, and is neither PIF-regulated nor light-responsive in the root.
•
This dynamic transcriptional module acts with the circadian clock to adjust the level of rhythmic gene expression. Sudden changes in either light or temperature adjust the equilibrium of HY5 and PIF bound to target promoters, altering the transcriptional response. In this way the expression of genes involved in the biosynthesis of MEP-derived plastidial isoprenoids, such as chlorophylls and carotenoids, can be modulated by external signals.
b. “Anti-rhythmic” regulation of DXS protein levels (WP2.3 2.5)
A straightforward interpretation of the morning-peaking RNA levels of the genes in the carotenoid pathway would be that carotenoid synthesis is also most active in the morning. A physiological advantage was previously proposed for this regulation, namely increased protection against predictable light stress, promptly from the start of the light period. However, TiMet’s analysis of multiple tocopherol and carotenoid compounds showed no rhythmic regulation under constant light conditions, presenting a paradox: RNA levels were clearly and co-ordinately rhythmic but the levels of the metabolic products were not. Our work to resolve this paradox revealed a previously-undescribed mode of “anti-rhythmic” protein regulation, with evidence that this regulation is important for plant growth.
Significant results:
• DXS and HDR proteins were shown to be rapidly degraded, in isolated chloroplasts, suggesting that the rhythmic expression of their RNAs would lead to a morning peak in protein levels.
•
• DXS degradation required Clp proteases in the chloroplast, mediated by a J-protein chaperone that likely targets misfolded forms of DXS.
•
• DXS protein level in wild-type plants changed very little over the normal light:dark cycle.
•
• Surprisingly, DXS protein levels did oscillate in transgenic plants bearing a constitutively-expressed, 35S:DXS transgene. In these plants, the protein level peaked in the evening.
•
These results suggested that the morning peak of rhythmic DXS RNA levels is normally counteracted by greater accumulation of DXS protein in the evening. The rhythmic RNA and ‘anti-rhythmic’ protein regulation results in the constant protein levels observed in wild-type plants.
• The growth defect of dxs knock-out mutant plants cannot be rescued with a constitutively-expressed 35S:DXS transgene (the resulting plants grew slowly). 35S:HDR and 35S:PSY constructs could not rescue hdr and psy mutant growth at all, resulting in a seedling-lethal, albino phenotype like dxs.
•
This result is consistent with the notion that constant levels of these proteins are important for normal plant growth. The tentative conclusion is that the clock-regulated RNA expression is essential to avoid a detrimental oscillation in enzyme levels. Presumably, the constant levels of enzymes are crucial to maintain constant levels of the final isoprenoid products, which are important for plant growth.
c. Regulation of DXR protein by the clock (WP2.3 2.5)
No changes in transcript levels were found for the gene encoding DXR, another rate-determining enzyme of the MEP pathway. However, levels of DXR protein (but not RNA transcript) were increased in toc1-1 mutant plants, which have a 20h biological clock due to the defect in the clock component TOC1. This molecular effect made the plants resistant to the DXR inhibitor compound, fosmidomycin, and suggested that the circadian clock was involved in regulating DXR.
Significant results:
• Several other clock mutants and a clock-affecting transgenic line also increased resistance to fosmidomycin, indicating that the result was not specific to toc1.
•
• Growth in 20h light:dark cycles restored the toc1 mutant to fosmidomycin sensitivity, as expected if the mutant’s phenotype was due to the mutant clock’s abnormal phase in 24h cycles.
•
• DXR protein levels were constant across the day in wild-type plants, and similar in 12h and 16h photoperiods. In toc1 mutants, DXR levels were higher at dawn, especially in 16h photoperiods, suggesting that the clock influences DXR protein stability. The DXS protein was not affected.
•
• DXR protein levels were shown to be regulated by Cpn60 chaperonins in wild-type plants, distinct from the mechanism regulating DXS.
•
d. Models of isoprenoid metabolism (WP2.3 2.4 2.5)
TiMet first developed metabolic pathway models of all currently known genes for enzymes in isoprenoid metabolism (MVA, MEP and associated pathways), annotated with the relevant Enzyme Catalogue numbers and linked to experimental data on the cognate Arabidopsis genes. These results were presented online in the AtIPD resource (Section 5a).
TiMet developed the first model of the MEP pathway of plastidial isoprenoid synthesis, integrating several mechanisms that influence the daily rhythms of the pathway. First, stimulation of the pathway flux by light through the supply of the substrate glyceraldehyde 3-phosphate (GAP) from photosynthesis. Second, feedback regulation of the key enzyme 1-deoxy-D-xylulose 5-phosphate synthase (DXS) by MEP-derived products. Third, circadian regulation of three steps in the pathway (DXS, HDR and consumption of ultimate products by other pathways). Regulation in sink and source tissues was distinguished.
Significant results:
• Enzymatic steps of the MEP, MVA and downstream isoprenoid pathways were annotated with the relevant Enzyme Catalogue numbers, AGI gene identifiers, and relevant proteomics data.
•
• The dynamic MEP model suggests that circadian regulation of the MEP pathway only slightly stimulates flux when plants are exposed to light during the day.
•
• Changes in GAP substrate availability from photosynthesis in source tissues are predicted to alter the MEP pathway flux much more strongly between day and night, compared to clock regulation.
•
• In sink tissue, circadian control might have a greater effect, comparable to the extent of substrate-level control, because substrate availability is less dependent on light.
•
• Jointly, light-dependent substrate and clock-modulated enzyme activities ensure that sufficient MEP product are available during the light phase of a day, when their demand is highest.
•
4. New understanding of plant growth.
a. Day/night balance of growth (WP1.1 1.2 2.5 3.1 3.3)
Growth in developmental biology is often taken to mean an increase in size, which is easily measured by non-destructive imaging. In TiMet, we focussed on growth as the conversion of carbon into cellular components. Growth in this context means biomass. 90% of fresh biomass in plants is water, which can be affected by many processes other than the metabolism of interest to us. Therefore the most relevant measure for us is dry biomass, which can only be measured destructively. Moreover, the composition of biomass in terms of cell wall, protein, etc., might in principle vary. We developed methodologies based on whole plant carbon balances, flux analysis to different organs and biomass components, including detailed analysis of amino acid pools and protein synthesis, and on analyses of ribosome abundance and polysome loading, that provide quantitative information about growth rates and mechanisms in the daytime and night.
Significant results:
• the overall rate of growth in a 24 h cycle is dependent on the net availability of carbon.
•
• the diurnal timing of growth is highly flexible, such that total growth in darkness can be as great as growth during the day.
•
• growth at night is determined by the amount of starch at dusk, the duration of the night and the rate of maintenance respiration,
•
• protein synthesis at night accounts for a large part of the total carbon and energy that is available from the degradation of starch,
•
• clock-dependent timing of starch degradation leads to stable changes in the sucrose level across most of the night,
•
• changes in sucrose content correlate with polysome loading (a proxy for the rate of protein synthesis),
•
• the daytime or night growth rate correlates with the corresponding content of sucrose, Tre6P and total amino acids,
•
• slow growth at night can lead to a delay before rapid growth is established on the following day
•
• in extreme cases, the delay occurs with signs of protein breakdown by a starvation response.
•
• carbon allocation is highly flexible during the 24 h cycle, consistent both with estimates from component reactions and with direct measurements of the flux to different biomass components,
•
• carbon allocation to different organs varies over plant development,
•
• the clock orchestrates ongoing rhythms of biochemical growth even in constant conditions,
•
• altered circadian timing of growth is correlated with effects on carbon metabolism.
•
These results led to a new, systems-level perspective, in which the quality, depth and breadth of data are sufficient to analyse the levels of central carbon metabolites (and increasingly, protein synthesis) as emerging from the balance of multiple influences upon the fluxes of supply and demand. This perspective was elaborated first on a daily and now on an hourly timescale.
b. Connecting clock outputs and the Framework Model (FM); WP3.1 3.3.
To represent the complex interactions uncovered by our results, and to provide a tool to further future understanding, TiMet built the first multi-scale, mathematical model of whole-plant growth for Arabidopsis thaliana. One component was to connect the model of the clock gene circuit to its best-characterise molecular outputs, the photoperiodic flowering pathway (including FKF1 and FT, see sections 2.b.i and 5e) and the light-regulated elongation pathway (including the PIF factors, see section 3.a.ii). The second was an integrated Framework Model (FM) that links gene network dynamics to the organ and whole plant levels, thus providing a means to understand the combined effects of circadian control and environmental signalling on carbon partitioning into starch, starch degradation, resource allocation and ultimately plant form. Validating the model depended on the rich data from TiMet.
Significant results:
• the clock-output model predicted new connections and dynamic features in the FT and PIF pathways, which were validated by experimental data,
•
• the clock-output model was extended to control hundreds of photoperiod-regulated PIF target genes, giving it genome-wide scope for the first time.
•
• the FM was built by integrating four models from the literature, in a modular fashion.
•
• the FM made accurate biomass predictions in reference conditions, with minimal recalibration, matching new data from Edinburgh and from Golm.
•
• the model was extended to include stochastic development, and resolved unanswered questions on the phenotype of a developmental mutant.
•
• simulations of the FM show how our data invalidate the earlier ‘overflow’ hypothesis for allocation of ‘spare’ carbon to starch, and require a flexible, conservative strategy for starch storage under different photoperiods,
•
• the modular model structure allows individual sub-models to be replaced and updated easily,
•
• a version of the model was developed in a user-friendly, graphical modelling software, to facilitate dissemination and re-use; online tutorials were held for a live, international audience.
•
• integrating the clock-output model allowed us to simulate the pleiotropic effects of a long-period clock mutant, matching its effects upon starch degradation, elongation growth, flowering and biomass growth.
•
• other measured carbon reserve pools are affected by the mutant (section 1d) and are required, together with the starch defect, to account quantitatively for the decreased growth phenotype of this mutant.
TiMet has therefore demonstrated that it is possible to establish a quantitative link between a change in the dynamics of a gene regulatory network (in this case, the circadian clock) and a complex emergent phenotype like whole-organism biomass, for the first time in any large eukaryote. Our results indicate that existing sub-models produced in different laboratories can usefully be combined, despite the variation in plant growth observed among laboratories, so long as each laboratory’s experiments are appropriately documented. Thus many laboratories can contribute to a community-based effort to develop and extend such models.
c. Methods for spatial analysis (WP3.1 3.3)
In parallel with our focus on biomass growth, two new technical developments allowed us to measure elongation growth, contributing to our understanding of hourly timescales.
i. Root elongation
Root extension growth is an easily-measurable growth parameter in young Arabidopsis plants, because it occurs in only one dimension, in a single organ that elongates without other movement. Imaging studies revealed a circadian oscillation in root extension rate, with a trough at ZT8-10, which depends on the evening complex. The growth pattern was modified by carbon supply: low carbon leads to an almost immediate inhibition of growth. The importance of clock-regulation of starch degradation for growth was demonstrated by showing that premature exhaustion of starch in the short period lhy cca1 double mutant leads to an inhibition of root extension growth at the end of the night, which is recovered by adding exogenous sucrose. The root extension assay is therefore a time-resolved, physiological marker for carbon availability in the whole plant.
ii. Rosette expansion
Rosette expansion growth is a far more complex process, involving multiple organs that both expand and also move, in three dimensions. TiMet developed a novel method for spatially- and temporally-resolved analysis of rosette growth, using a newly-developed, commercial light-field camera. The method uses only one camera, greatly simplifying the process compared to competing methods. The equipment, with a significant development of image processing and analytical software, allows reconstruction of a 3D image of an Arabidopsis rosette at each timepoint, and hence corrects for leaf movement, allowing precise quantification of expansion growth in a freely moving plant. It is now being applied to test the range of samples and conditions of interest to TiMet.
5. New research methods and extensions to unexpected areas.
The creativity and reach of our partners allowed the project to return exciting results in unexpected areas relevant to the project.
a. Data infrastructure and analysis methods (WP1, WP2.4 2.5)
TiMet’s data management infrastructure reflected the diversity of large-scale experimental methods and modelling methods used in the project.
• A comprehensive and interactive Arabidopsis Isoprenoid Pathway Database (AtIPD) was constructed (http://www.atipd.ethz.ch/) which includes TiMet’s metabolic pathway models of the isoprenoid pathways, linked to the relevant Enzyme Catalogue numbers, data on the cognate Arabidopsis genes and subcellular localisation. The isoprenoid pathway models can be downloaded in several formats including for the Cell Designer modelling tool.
•
• A new relational database, TiMet-DB, was constructed to allow powerful queries across multiple, genome-scale data sets, including the TiMet transcriptome and proteome data. Analysis routines can be added to the database, organising both primary and secondary data.
•
• The BioDare online resource (www.biodare.ed.ac.uk) was extended with the most powerful and specific tools available to retrieve desired data from among hundreds of similar experiments, allowing sharing among collaborators and public dissemination. From two methods initially, BioDare’s functions were extended to 6 mathematical methods for online analysis of circadian rhythm data. New visualisation methods enhanced data display. RNA and metabolite timeseries data from several labs were stored in the repository.
•
• The PlaSMo model repository (www.plasmo.ed.ac.uk) was updated and hosts public dissemination of TiMet’s mathematical models, including the Framework Model.
•
b. Models for surface-active enzymes (WP2.4)
TiMet’s work on starch differs from most biochemistry in one important respect: the starch granule is a quasi-crystal. Enzymes that metabolize starch do not do so in solution, but on the granule surface. TiMet first considered how entropy affects the activities of enzymes that act on carbohydrate polymers like starch. Applying these ideas to the starch surface, however, prompted TiMEt researchers to develop a much more general, theoretical framework that describes surface-active enzymatic reactions. The new theory derives rate laws analogous to the classical Michaelis-Menten rate law for enzymes acting in solution. This development was not anticipated at the start of the project but it should be broadly applicable to many enzymes that function on solid surfaces, such as the surface of the starch granule.
c. Economic analysis (WP3.3)
Many results suggest that plants partition their resources in an optimal way, which resembles an economic optimisation of profit. This concept allowed a novel, conceptual approach to understand plant growth, which we expect to have broad relevance. TiMet considers the partitioning of growth between day and night, and of carbon between sugar for use and starch for storage. During the day, energy is freely available from photosynthesis, whereas at night, energy has to be remobilised from starch and other stores and is thus more expensive. However, the process of growth also requires machinery, which must be built using considerable resources. Thus it may be ‘worth’ growing during the night, even if energy costs are higher, rather than letting the valuable machinery lie idle. The analogy is with a capital-intensive factory, where it is advantageous to run a night-shift even if the wages during the night shift are higher than during the day. In preliminary studies, these economic principles were embodied in a mathematical model, in order to find the optimal balance of day- and night-time growth, depending on the energy invested in the growth machinery. Strikingly, the model’s predicted optimal strategies were the opposite of the experimental observations, indicating that other factors are important in the plant’s strategy. The optimality-based approach promises to guide future understanding of plant growth regulation, by revealing the factors that allow the model to match the data.
d. Translational Coincidence (WP2.4)
Data and insight from four TiMet partners contributed to discover an operating principle for plant gene expression, which we term “translational coincidence”. This molecular mechanism contributes to answer a fundamental question posed in our original application, namely how daily rhythms mediated by the circadian clock contribute to seasonal adjustment of plant metabolism under different photoperiods.
The concept was originally developed from TiMet observations of changes of protein abundance under different photoperiods and of different rates of translation in day and night (sections 1.c 4.a). The translational coincidence mechanism predicts a change in the levels of stable proteins in different photoperiods, if their translation rate in light is greater than in darkness, and if the transcript levels are rhythmic and controlled by a circadian clock. In Arabidopsis, the proteins encoded by rhythmic RNAs with peak levels in the evening were predicted to become more abundant under long photoperiods, whereas proteins from morning-peaking RNAs would become relatively less abundant. This prediction, along with several more specific, secondary predictions, was validated by hundreds of proteins in TiMet’s quantitative proteomics data and also by data from two independent projects. This insight may contribute to resolve a longstanding paradox: many rhythmic RNAs encode very stable proteins that do not change in abundance over the course of a day. Translational coincidence would allow these daily rhythms to control seasonal changes in protein abundance, where the time of the daily RNA peak determines the season of peak protein expression. Preliminary results suggest that this mechanism may operate generally, across all photosynthetic organisms, giving it far-reaching implications.
e. The Tre6P signal controls flowering as well as vegetative growth (WP2.2)
During the project, we discovered that the sugar signalling molecule Trehalose-6-Phosphate (Tre6P; see section 2.c.ii) interacts with another important clock output, namely the floral induction pathway. Plants engineered to increase T6P were known to be early flowering, whereas transformants with reduced Tre6P were late flowering. Thus the Tre6P signal might mediate the longstanding suggestion that sugar status can control flowering time, which previously had no known molecular mechanism. Moreover, it was unknown how this effect interacted with other known signalling pathways that regulate floral initiation in response to environmental signals like photoperiod and temperature. By analysing expression of the clock output genes GI and CO, and the floral mediator FT, in lines with altered Tre6P, we showed that Tre6P is not required for clock-dependent rise in GI and CO transcripts, but is required for the increase in FT transcript. Genetic analysis that tested the interdependence of the GI- and Tre6P-dependent regulation, led to the suggestion that Tre6P mediates a carbohydrate-dependent modulation of flowering, sensitising flowering to photoperiod when sucrose is high, and attenuating the response when sucrose is low. Thus the plant’s sugar status signalled via Tre6P interacts with circadian clock outputs to control the developmental transition to flowering, analogous to its effect on starch degradation to control vegetative growth.
f. Towards natural regimes: natural light, double-dawn and climate change scenarios (WP3.3)
Almost all previous experiments performed on the clock and on starch metabolism use a simple step light regime, which changes instantaneously from darkness to full light intensity and back. However, in natural light regimes the light intensity changes gradually. TiMet results led us to investigate how the clock is entrained by light in natural regimes and how this affects carbohydrate metabolism. Clock gene RNA levels, starch and key metabolite levels were tested in plants grown in regimes providing periods of high light (adequate for photosynthesis) and very low light (able to entrain the clock), in natural light regimes, and in regimes in which light levels fall abruptly at different times of day. Clock dynamics were modified in an unforeseen manner in natural light regimes, especially during the night, when a secondary feature of dusk and evening complex genes was greatly enhanced. The natural light also delayed the phase of the clock, and the timing of starch degradation. We propose that this delay will allow plants to maintain a sugar supply from starch in natural light environments where the light intensity rises gradually, such that ‘circadian dawn’ occurs before ‘photosynthetic dawn’.
Natural conditions are expected to change under predicted global climate scenarios, which involve increases of temperature and CO2 as well as water shortage in many areas of the globe. A large-scale, multifactorial experiment extended TiMet’s work, in collaboration with the European Science Foundation MOMEVIP project, to test the response of Arabidopsis wild type and isoprenoid mutants to increased temperature, CO2 and drought conditions. Detailed phenotypic, physiological and volatile data were acquired by the partners, while TiMet performed transcriptome analysis and ongoing proteome measurements. Among the genome-wide analysis, a striking observation was the loss of transcript fluctuations for three core clock genes between the end of the day (EOD) and the end of the night (EON) in the most severe, multiple-stress condition. These effects, however, disappeared when plants were re watered.
Potential Impact:
The TiMet project pioneered the dynamic, systems-level analysis of combined metabolic and gene-regulatory networks and their multiple links to growth. The project aimed to develop fundamental understanding of this important system, as a test case for using model-guided, mechanistic tools to analyse biological regulation across scales. This was only possible in the laboratory model species, Arabidopsis thaliana. Establishing the toolset and proof-of-principle in this species, however, will now justify and direct future work to apply this systems biology approach to other biological processes, and to other species including the best-known crops.
The partners maintained a close review of the emerging results from their work with a view to protecting the results, in line with the consortium agreement and as appropriate for a fundamental research project of this nature. No patent filings or other intellectual property protection were initiated by any partners from work on the project.
To engage and inform the stakeholders who will be impacted by our work in future, TiMet attended or organised four meetings with a broad range of stakeholders, in addition to the ongoing activities of individual partners in this and related research areas:
• Presentation to stakeholders at the FESPB and European Plant Science Organisation Plant Biology Congress, Freiburg, 2012, with over 1000 participants from 60 countries.
• Attendance at events of the EU COST action PlantEngine, 2013.
• Organisation of a half-day ‘research programming’ workshop in Barcelona in 2013 for the TiMet Stakeholder Platform, attended by 7 EU FP6 or FP7 projects in Plant Systems Biology, along with 7 companies and other organisations, in addition to all TiMet partners.
• Publication of a professionally-written “Success Story” for DG Research & Innovation, 2014.
• Presentations to the TiMet Stakeholder Platform attending a half-day workshop, organised before the TiMet final meeting, in Dusseldorf 2015.
All partners participated actively in dissemination events in many forums, communicating their results to the scientific community through conferences, seminar and publications.
Public dissemination was undertaken in events ranging from science fairs to public lectures, in addition to the public website www.timing-metabolism.eu and micro-blogging on the Twitter platform (@TiMet_project). The website was redesigned around the mid-term of the project to adopt a ‘weblog’ (aka blog) format, with short articles on the project’s work and related items of interest, organised in multiple themes, allowing frequent updates by all partners.
A detailed list of these activities is provided separately.
List of Websites:
http://www.timing-metabolism.eu