Diurnal changes of polysome loading track sucrose content in the rosette of wild-type arabidopsis and the starchless pgm mutant.

Growth is driven by newly fixed carbon in the light, but at night it depends on reserves, like starch, that are laid down in the light. Unless plants coordinate their growth with diurnal changes in the carbon supply, they will experience acute carbon starvation during the night. Protein synthesis represents a major component of cellular growth. Polysome loading was investigated during the diurnal cycle, an extended night, and low CO2 in Arabidopsis (Arabidopsis thaliana) Columbia (Col-0) and in the starchless phosphoglucomutase (pgm) mutant. In Col-0, polysome loading was 60% to 70% in the light, 40% to 45% for much of the night, and less than 20% in an extended night, while in pgm, it fell to less than 25% early in the night. Quantification of ribosomal RNA species using quantitative reverse transcription-polymerase chain reaction revealed that polysome loading remained high for much of the night in the cytosol, was strongly light dependent in the plastid, and was always high in mitochondria. The rosette sucrose content correlated with overall and with cytosolic polysome loading. Ribosome abundance did not show significant diurnal changes. However, compared with Col-0, pgm had decreased and increased abundance of plastidic and mitochondrial ribosomes, respectively. Incorporation of label from (13)CO2 into protein confirmed that protein synthesis continues at a diminished rate in the dark. Modeling revealed that a decrease in polysome loading at night is required to balance protein synthesis with the availability of carbon from starch breakdown. Costs are also reduced by using amino acids that accumulated in the previous light period. These results uncover a tight coordination of protein synthesis with the momentary supply of carbon.


Introduction
Protein synthesis occurs via recruitment of ribosomes to mRNA to form polysomes (Bailey-Serres et al., 2009). It represents a major component of the total ATP consumption in animal and plant cells (Hachiya et al, 2007;Pace and Manahan, 2007;Proud, 2007;Piques et al., 2009;Raven 2012). For each amino acid added to the growing peptide chain, two ATP are consumed in aminoacyl-tRNA synthesis and two in peptide bond synthesis. The actual costs are higher due to copy reading, and because many proteins are synthesised as longer polypeptides and then trimmed to their final size.
Energy is also required to synthesise amino acids. Conversion of nitrate to amino acids requires the equivalent of about 5 ATP and, on average, 2.8 C per amino acid (Penning de Vries, 1975;Hachiya et al, 2007, Amthor et al., 2010. Protein synthesis also carries substantial indirect costs. Mature ribosomes contain 4 ribosomal RNA (rRNA) species (typically 25S, 18S, 5.8S and 5S) and approximately 80 ribosomal proteins (Bailey-Serres et al., 2009). Ribosomal RNA and ribosomal proteins represent >80% and 30-50% of the total RNA and protein, respectively, in a growing yeast cell (Warner, 1999;Perry, 2007). Ribosome biogenesis involves synthesis of a large ca. 45S rRNA precursor that is processed to generate the mature rRNA species and synthesis of ribosomal proteins, and their stepwise assembly into the large and small ribosome subunits in a process that requires about 200 ancillary proteins (Houseley and Tollervey 2009).
Ribosome synthesis and ribosome loading are regulated by the universal nutrientsignalling TOR (Target of Rapamycin) pathway in animals and yeast (Wullschleger et al., 2006;Mayer andGrummt, 2006, Ma andBlenis, 2009). Inducible inhibition of TOR expression revealed that TOR is also a major regulator of metabolism and growth in plants (Caldana et al., 2012). Synthesis of the 45S rRNA precursor in Arabidopsis is regulated by the kinase domain of TOR (Ren et al., 2011) and Arabidopsis mutants with strongly decreased TOR expression show a small decrease in polysome loading (Deprost et al., 2007). In yeast and animals, TOR regulates polysome loading via a signal cascade initiated by the AMP-dependent protein kinase or SNF1, leading to phosphorylation of the ribosomal protein S6 and of the initiation factor eIF4E-binding protein eIF4BP and elongation factor eEF2 (Ma and Blenis 2009). Phosphorylation of ribosomal protein S6 is implicated in stress signalling in plants (Scharf and Nover 1982;Williams et al., 2003, Mahfouz et al., 2006. The daily alternation between light and darkness is one of the most pervading environmental changes experienced by plants. In the light, photosynthetic electron transport and photophosphorylation deliver ATP and NAD(P)H, providing energy to assimilate CO 2 into carbohydrates and nutrients like nitrate and ammonium into amino acids. In the dark, carbohydrates and other C-containing storage metabolites are catabolized to generate C-skeletons, NAD(P)H and ATP. This involves energy costs, including loss of free energy during the turnover and respiration of C reserves.
Starch is the major C reserve in many species (Geiger et al., 2000;Smith and Stitt, 2007; Stitt and Zeeman 2012). Arabidopsis mutants impaired in starch synthesis or degradation show strongly reduced growth except in continuous light or very long days (Caspar et al,. 1985(Caspar et al,. , 1991. In wild-type Arabidopsis, growth is rapidly inhibited when starch is exhausted and this inhibition is not immediately reversed when C becomes available again (Gibon et al. 2004a;Smith andStitt, 2007, Yazdanbakhsh et al., 2011). The risk of acute C starvation is minimised by regulating the rate of starch degradation; this occurs in a near-linear manner such that most but not all of the starch is exhausted at dawn (Smith and Stitt 2007;Stitt and Zeeman 2012). This pattern of starch turnover is retained across a wide range of growth conditions (Chatterton and Silvius 1979;1980;reviewed in Smith and Stitt 2007;Stitt and Zeeman 2012). The rate of starch degradation is set such that starch is almost exhausted at dawn as anticipated by the biological clock (Graf et al. 2010). This allows the rate of degradation to be immediately adjusted to sudden and unpredictable changes in the amount of starch at dusk (Lu et al,. 2005;Graf et al., 2010) or night temperature (Pyl et al., 2012).
This sophisticated regulation of photosynthate allocation needs to be accompanied by coordinated changes in the rate of growth (Stitt and Zeeman, 2012). A decrease in the C supply at some time during the diurnal cycle due to the alternation of light and darkness, changes in the growth conditions or sudden unpredictable changes like, for example, shading or changes in the rate of starch degradation will result in acute C starvation unless there is a concomitant decrease in the rate of C utilization.
During diurnal cycles there are dynamic changes in the rate of leaf and root extension growth (Schmundt et al., 1998;Walter et al., 2009;Poire et al., 2010;Yazdanbakhsh et al., 2011), which are modified in response to short-and long-term changes in the C supply (Wiese et al., 2007;Gibon et al., 2009;Yazdanbakhsh et al., 2011;Pantin et al., 2011;Kjaer et al., 2012). However, extension growth is driven by water uptake and vacuole expansion. Such measurements do not provide information about the timing of the biosynthesis of cellular components.
Protein synthesis provides an experimentally tractable process (Rudra and Warner, 2004) to study the relation between the C supply and the rate of cellular growth. Polysome loading in Arabidopsis rosettes increases between the end of the night and 2 h into the photoperiod (Piques et al., 2009) and decreases slightly when plants are darkened for 1 h in the middle of the light period (Juntawong and Bailey-Serres, 2012). However, it is not known whether polysome loading or ribosome abundance change in the remainder of the diurnal cycle, or whether any such changes are in response to light, C fixation or other inputs. An additional complication is introduced by subcellular compartmentation. Plant cells contain considerable amounts of ribosomes in their plastids (Detchon and Possingham, 1972;Dean and Leach, 1982). Plastid translation is especially dependent on light (Deng and Gruissem 1987;Marin-Navarro et al., 2007). The cvtosol, plastid and mitochondria have contrasting phosphorylation potentials that respond differently to illumination and darkening (Stitt et al., 1982;1983;Gardeström and Wigge, 1988;Igamberdiev and Gardeström 2003).
The following experiments investigate polysome loading and ribosome abundance during diurnal cycles in wild-type Arabidopsis and the starchless pgm mutant. We show that overall polysome loading changes dynamically during diurnal cycles, closely tracking sucrose levels. Cytosolic polysome loading responds mainly to changes in sucrose, plastidic polysome loading shows a strong dependence on light, and mitochondrial polysome loading remains high throughout the diurnal cycle. While ribosome number remains similar throughout the diurnal cycle and for at least 8 hours into an extended night, the abundance of plastidic and mitochondrial ribosomes is modified in the starchless pgm mutant, indicating long-term adjustment of organelle ribosome number to sugars. This information about ribosome abundance and loading into polysomes is then used to model the rate of protein synthesis and the associated costs and compare them with the availability of C at different times in the diurnal cycle.

Changes of metabolites during a diurnal cycle in wild-type Col-0 and the starchless pgm mutant
Wild-type Col-0 and the starchless pgm mutant were grown in a 12 h light / 12 h dark cycle. After 5 weeks, rosettes were harvested at the end of the night, after 0. 25, 0.5, 1, 2, 4, 8, 12 h in the light, and after 0.25, 0.5, 1, 2, 4, 8 and 12 h in the dark. On the following day the light was not turned on in the morning, and further sets of plants were harvested 0.5, 1, 2, 4 and 8 h into the extended night. The diurnal changes of carbohydrates were analyzed in biological triplicates to provide an internal baseline for comparison with changes in ribosome loading and abundance ( Fig. 1, for original data see Supplemental Table SII). Whilst the results resemble earlier studies (e.g. Caspar 1985;Smith, 2004;Gibon et al., 2004a;2006;Bläsing et al., 2005;Graf et al., 2010), the increased density of time points provides additional information, especially during transitions between light and darkness.
In wild-type Col-0, starch accumulated in a near-linear manner in the light and decreased in a near-linear manner in the dark, with almost all the starch being exhausted by the end of the night (Fig. 1A). Sucrose (Fig. 1B) rose to a maximum after 30 min, remained high for the remainder of the light period, decreased to a transient minimum 15-30 min after darkening, partially recovered after 1-2 h, declined slightly during the remainder of the night and decreased by 60% during the first 4 h of the extended night. Glucose and fructose increased more gradually than sucrose at the start of the light period, decreased These metabolites showed very different kinetics in pgm, both with respect to the timing and magnitude. The scales of the y-axis for the Col-0 and pgm displays are different; to aid comparison, the Col-0 response is indicated as a dotted line in the pgm displays. First, as expected, starch is effectively absent in pgm (Fig. 1E). Second, following illumination, sucrose rose to very high levels, with a peak after 1 h followed by a 50% decrease during the remainder of the light period (Fig. 1F). Glucose and fructose also rose to very high levels, but more slowly than sucrose, reaching maximum values after 4-8 h in the light and remaining high until the end of the light period (Fig. 1G, Supplemental Table II). The total amount of carbon in reducing sugars was >2-fold higher than that in sucrose. After darkening, glucose and fructose decreased rapidly within the first 15-30 min, whilst sucrose decreased gradually over the first 4 h of the night. In contrast to wild-type plants, glucose-6-phosphate showed a strong increase during the light period, and a rapid decrease after darkening (Fig. 1H). The levels of sucrose, reducing sugars and glucose-6phosphate in pgm at the end of the night resemble those in Col-0 in an extended night.
The transient decrease of sucrose and glucose-6-phosphate after darkening in wild-type Col0 resembles the response seen in earlier studies of spinach, barley (Stitt et al., 1985).
One explanation for this transient decrease would be a delay before starch degradation commences after a sudden shift from light to darkness. The absence of a transient decrease in pgm is consistent with this possibility. To provide additional evidence, maltose levels were analyzed (Fig. 2). Maltose is an intermediate of starch degradation in leaves (Niittylä et al., 2004;Stitt and Zeeman 2012). Maltose levels were relatively high at the end of the night, remained high for the first 30 min after illumination, declined to low levels for most of the light period, remained low for the first 15 min in the dark, rose progressively at 30, 60 and 120 min after darkening, and decreased gradually during the remainder of the night. These results indicate there is a lag until starch degradation is activated and inhibited after sudden darkening and sudden re-illumination, respectively.

Changes of polysome loading during a diurnal cycle in wild-type Col-0
Material from the same set of biological triplicates was individually subjected to polysome density gradient centrifugation. Examples of typical gradients from material harvested at the end of the night and after 2 h illumination are provided in Supplemental  Fig. S1. The distribution of RNA is monitored via absorption at 254 nm (A 254 ). As the vast majority of the RNA in the gradients is rRNA (Raven, 2012), A 254 largely reflects the distribution of ribosomes. At the end of the night, the majority of the RNA was present at the top of the gradient, corresponding to free ribosomes (non-polysome fraction, NPS). The rest was present in an intermediary small polysome fraction (SPS, corresponding to polysomes with 2-4 ribosomes) and a large polysome fraction (LPS, corresponding to polysomes with 5 or more ribosomes). In the light, there was a large decrease of absorbance in the NPS fraction, and a large increase in the LPS fraction, reflecting an increase in the proportion of ribosomes that are loaded into polysomes. the night, about 40% of the RNA was in polysomes (SPS plus LPS), and the remaining 60% in the NPS fraction. After illumination, the fraction in polysomes rose to about 62% and 67% after 30 min and 1 h and remained high for the remainder of the light period.
After darkening, there was a rapid transient decrease of polysome loading (SPS + LPS) to <40%, followed by a partial recovery to about 50% during the first part of the night, a slight decline to about 40% at the end of the night and a further decrease to <20% when the night is extended. The main features of this response were seen in five independent experiments performed over a period of 3 years (see later).

Changes of polysome loading during a diurnal cycle in the starchless pgm mutant
Similar measurements were carried out with the starchless pgm mutant (Fig. 3B).
Polysome loading was low (about 25%) at the end of the night. It rose gradually over the next 2-4 h to a value of about 67%, remained high until the end of the light period, decreased gradually during the first 4 h of the night to 20-25%, and remained at this low value in the extended night. As in wild-type Col-0, the polysome loading tracked the sucrose content, with low values at the end of the night, a gradual rise in the light, and a gradual decrease during the first 4 h of the night. Differences in the response in Col-0 and pgm are highlighted in Figs. 3C-D, which compare the responses at the start of the light period and the start of the night. In pgm, ribosome loading into polysomes started from a lower value and rose more slowly after illumination (Fig. 3C) while at the start of the night polysome loading in pgm did not show a transient minimum and partial recovery, but instead fell gradually to a lower value than in Col-0 after 4 h darkness (Fig. 3D).

Comparison of polysome loading and metabolite levels
Visual inspection of Fig.1  Spearmans rank analysis yielded an even higher correlation (Rs = 0.82, p = 3x10 -9 ).
Visual inspection indicates that polysome loading is not so strongly correlated with the other measured metabolites (see Supplemental Fig. S2).
The strong relation between polysome loading and sucrose was checked by performing partial correlation analysis. This statistical approach analyzes the data matrix to exclude secondary correlations (Fig. 4B). Polysome loading was significantly correlated with sucrose (R = 0.69, p = 2x10 -5 ; note that the p-value is slightly lower than in a simple regression analysis due to correction for multiple testing) and weakly with starch (R = 0.46, p = 0.01) but not with any other measured metabolite.

Changes of loading of cytosolic, plastidic and mitochondrial ribosomes into polysomes at dusk and during the night
To resolve the responses of polysomes in the cytosol, plastid and mitochondria we performed a separate experiment in which we quantified the abundance of cytosolic, plastidic and mitochondrial 18/16S ribosomal RNA (rRNA) in the different density gradient fractions in wild-type Col-0 at the end of the day and several times during the night (Fig. 5, see Supplemental Table SIII for the original data). rRNA abundance provides a proxy for ribosome number. To allow absolute quantification, eight external standards were added before preparing RNA. The rRNA species were determined by qRT-PCR, using specific primers for the cytosolic, plastidic and mitochondrial 18/16S rRNA (see Supplemental Table S1). Ct values were corrected to absolute concentrations, using the external standards as a calibration curve (see Piques et al., 2009). In absolute terms, cytosolic, plastidic and mitochondrial ribosomes account for about 55, 45 and 2% of the total ribosomes (Piques et al., 2009 and below). Piques et al. (2009) showed that similar estimates of polysome loading are obtained using A 254 and by summing the rRNA species in the different fractions in a polysome gradient.
In the present study, the changes in polysome loading obtained by determining A 254 again resembled those obtained by summing the abundance of cytosolic, plastidic and mitochondrial rRNA (Fig. 5A).
Cytosolic, plastidic and mitochondrial ribosomes all showed a high loading at the end of the day (Fig. 5B). After darkening, cytosolic ribosome loading showed a small transient decrease at 30 min, a partial recovery, and declined towards the end of the night. Plastidic ribosome loading decreased strongly after 30 min dark and remained low for the remainder of the night. Mitochondrial ribosome loading remained high throughout the night. Whilst cytosolic rRNA, was always >3-fold higher in the LPS fraction than in the SPS fraction, the proportion of plastidic and mitochondrial rRNA in the LPS fraction was not much higher than that in the SPS fraction. In the dark, less plastidic rRNA was found in the LPS than the SPS fraction.

Changes of polysome loading and carbohydrates after illumination at subcompensation point and ambient CO 2 levels
It is possible that light leads to increased polysome loading independently of any changes in CO 2 fixation and carbohydrate levels. In particular, light is known to activate translation in chloroplasts (Marín-Navarro et al., 2007). To separate the effects of light and CO 2 fixation, we performed two further experiments in which wild-type Col-0 was  Table SIV). In 50 ppm CO 2 photosynthesis is prevented and there is even CO 2 release, whereas at 480 ppm CO 2 there is rapid photosynthesis and carbohydrate synthesis. The response of global gene expression to this increase of CO 2 closely resembles that after re-addition of sucrose to seedlings (Osuna et al., 2007).
In both experiments, sub-compensation point CO 2 completely suppressed the increase of starch, sucrose and reducing sugars that normally occurs after illumination (Fig. 6A).
Indeed, the levels of these metabolites decreased further, because the plants were exposed to an additional period of time in which there was no photosynthesis. In this particular experiment, slightly more starch remained at the end of the night and sucrose was slightly higher than in the experiments of Fig. 1 and other published studies (Gibon et al., 2004a;Bläsing et al., 2005, Usadel et al., 2008. Overall polysome loading assessed by A 254 increased from about 40% in the dark to 50 % after illumination at sub-compensation point CO 2 , and >60% after illumination at 480 ppm CO 2 (Fig. 6B). Addition of low concentrations of sucrose to C-starved seedlings also led within 30 min to an increase in overall polysome loading (Supplemental Fig. S3).

Response of the loading of cytosolic, plastidic and mitochondrial ribosomes into polysomes after illumination at sub-compensation point and ambient CO 2 levels
To resolve responses of translation in the cytosol, plastid and mitochondria, we quantified the abundance of cytosolic, plastidic and mitochondrial 18/16S rRNA in the different density gradient fractions at the end of the night, and after illumination for 2 h in the presence of sub-compensation point or ambient CO 2. The changes in polysome loading obtained by summing the cytosolic, plastidic and mitochondrial ribosomes in each fraction ( Fig. 6C) resembled those obtained by A 254 (Fig. 6B).
Cytosolic polysome loading hardly changed after illumination for 2 h at subcompensation point CO 2 , compared to the end of night (Fig. 6D). It rose strongly and significantly after 2 h illumination at ambient CO 2 (p = 0.016 compared to the end of the night, and p = 0.002 compared to illumination in low CO 2 ) (Fig. 6D). Plastid polysome loading increased strongly and significantly after illumination for 2 h at subcompensation point CO 2 (p = 0.004) and did not increase further in ambient CO 2 (Fig.   6D). Loading of mitochondrial ribosomes was high at the end of the night and in the light at sub-compensation point CO 2 , and showed a slight but non-significant decrease in the light in ambient CO 2 (Fig. 6D). In the dark a large proportion of the plastidic ribosomes were present in the SPS fraction compared to the LPS fraction.

Meta-analysis of diurnal changes in polysome loading
The data sets from the preceding experiments were combined with further data to examine the reproducibility of the diurnal response of overall polysome loading in five studies conducted over a time span of three years (Supplemental Fig. S4). The analysis revealed a remarkable reproducibility including the 2-fold rise after illumination, the transient trough 15-30 min after darkening, the subsequent recovery, the maintenance of relatively high polysome loading until the end of the night, and the decrease of polysome loading in an extended night (Supplemental Fig. S4A). This meta-analysis also revealed that the correlation between rosette sucrose content and overall polysome loading noted in the experiment of  Table SII for the original data). To decrease the analytic noise inherent in qRT-PCR measurements, three technical replicates were included for each biological sample. Cytosolic 18S rRNA abundance was similar in wild-type Col-0 and pgm and did not show any significant changes during the diurnal cycle and the extended night (Fig. 8A).
Plastidic 16S rRNA showed a slight non-significant increase during light period in Col-0 ( Fig. 8B). It was consistently lower in pgm than in wild-type Col-0. A reverse picture emerged for the mitochondria, where pgm consistently contained more mitochondrial 18S rRNA than wild-type Col-0 (Fig. 8C). The changes in pgm compared to Col-0 were significant (0.03 and <0.001) for plastidic and mitochondrial rRNA, respectively, using either the Holm-Sidak or the Tukey test). Thus, while there are no large diurnal changes of cytosolic, plastidic or mitochondrial ribosome number in either genotype, there are differences in the absolute amounts between Col-0 and pgm, with pgm containing slightly less plastidic rRNA and considerably more mitochondrial rRNA.

Comparison of changes in rRNA abundance and abundance of transcripts for ribosomal proteins
We mined public domain expression data for information about diurnal changes of transcripts that encode cytosolic, plastidic and mitochondrial ribosomal proteins (Usadel Transcripts for the vast majority of cytosolic ribosomal proteins were induced by sugars. They increased in the light and decreased in the night in Col-0 and showed more pronounced diurnal changes in pgm (Supplemental Figs. S6A and S7). A similar but even more pronounced response was found for transcripts encoding BRIX proteins and Nucleolin-1, which are involved in ribosome assembly (Supplemental Fig. S9). However, these diurnal changes in transcripts do not result in significant diurnal changes in cytosolic or mitochondrial ribosome abundance in Col-0 or pgm, as assessed by rRNA abundance (Fig. 8) Transcripts for mitochondrial ribosomal proteins showed a similar pattern to the cytosolic ribosomal proteins, with an increase in the light period in Col-0 that was accentuated in pgm (Supplemental Fig. S6C). The latter corresponds to the increase in mitochondrial rRNA in the starchless pgm mutant (Fig. 8).

Modeling the rate of protein synthesis and associated costs
The finding that polysome loading is positively correlated with sucrose content is understandable, as protein synthesis is an energy-intensive process (see Introduction).
However, the analyses presented so far do not provide any information about the quantitative relationship between the availability of C at different times in the diurnal cycle and the costs of protein synthesis. In particular, they do not reveal whether the decrease in polysome loading in the night is necessary to balance protein synthesis with the rate of starch mobilization.  Table SV. Costs were calculated as μ atom C g -1 FW h -1 ; parameters used to interconvert ATP and C are also summarized in Supplemental Table SV. The results are summarized in Fig. 9. Comparison of the modeled costs with the measured rates of starch breakdown and respiration leads to three predictions. First, the observed decrease in polysome loading at night is required to balance protein synthesis with the availability of C. Starch represents >80% of the C stored in an Arabidopsis rosette (Gibon et al., 2009;Pyl et al., 2012). If the rate of protein synthesis in the light were to be maintained at night, 25-33% of the starch and 36-47% of the measured respiration would be required to supply ATP for amino acid activation and peptide bond formation. This is unrealistic, as C and energy will be required for the synthesis of other cellular components and for maintenance. If the measured level of polysome loading at night is used as an input, less than 20% of the starch and 25% of the respiration is required to provide ATP for amino acid activation and peptide bond synthesis. Second, the rosette does not contain enough starch to support the synthesis of all the amino acids that are used for protein synthesis at night. The estimated full costs in the night are equivalent to about 73-84% of the available starch, and are similar to or higher than the rate of respiration. This indicates that a substantial proportion of the amino acids that are used at night may be accumulated during the preceding day. Amino acids often accumulate in the day and decrease at night in leaves (Pate, 1989;Morot-Gaudry et al., 2001). Supplemental Fig. S10 compares the modeled rate of amino acid incorporation into protein with 5 studies of diurnal amino acid turnover in Col-0 growing in a 12h light/12 h dark cycle. While there is variation between experiments, this comparison indicates that up to half the amino acids that are used for protein synthesis at night are synthesized in the preceding light period. Third, the cost of protein synthesis in the light (13-18 μ atom C g -1 FW h -1 ) is equivalent to 18-21% of the total fixed C (85 μ mol CO 2 g -1 FW h -1 , Supplemental Table SV). This value will be an underestimate because some of the amino acids that are used at night are synthesized in the preceding light period.

Estimation of the rate of protein synthesis rates from 13 CO 2 incorporation
Our model predicts that protein synthesis continues at a substantial rate during the night.
To test this prediction, we monitored incorporation of 13 CO 2 into protein. Whole plants were transferred before dawn into a chamber that was supplied with a stream of 480 ppm 13 CO 2 . Some plants were harvested before transfer to measure 13 C natural abundance, and others were harvested at the end of the day or the end of the night. Labeling was started just before dawn, when most endogenous pools are at their diurnal minimum. Starch represents about 80% of the total C reserve in Arabidopsis (Gibon et al., 2009) and is almost completely depleted at the end of the night (Fig. 1A). Other metabolites including sugars (see Fig. 1B) and amino acids (Gibon et al., 2009) are also at a minimum at dawn.
This experimental design ensures that starch and other C reserves are built up in the light period using newly fixed C, providing a highly-enriched source of C for metabolism at night. The supply of 13 CO 2 was maintained throughout the night to avoid dilution of these internal pools by CO 2 fixed by PEP carboxylase. Total protein was extracted and chemically hydrolyzed to release amino acids for analysis by GC-MS. The mass shift resulting from incorporation of one or more atoms of 13 C allows identification of the 12 C species and the various 13 C isotopomers (Szecowka et al., 2013). Data was obtained for incorporation in the night from the increment in enrichment between the end of the day and the end of the night. It should be noted that the estimated rates are relative, because they are not corrected for enrichment in the precursor pools of free amino acids.
When 13 C enrichment is averaged across all amino acids, it increased on average by 1.4% per h in the light, and 0.4% per h in the night. This indicates that the rate of protein synthesis is about 3-fold lower in the dark than in the light period. Overall, about 60% of the protein synthesis occurred in the light period, and 40% in the night. For comparison, polysome loading measured in this plant material was 60% in the light, and 40% in the dark. This is smaller than the inhibition of protein synthesis estimated from 13 C incorporation (see Discussion).
One possible explanation for the decreased rate of 13 C incorporation into protein in the dark might be recycling of unlabelled amino acids released by protein degradation. If recycling were leading to an underestimation of the rate of protein synthesis in the dark it should have an especially marked effect on minor amino acids, because they are more likely to be recycled without mixing with C from central metabolic pools. We inspected the responses for each individual amino acid (Supplemental Table VI polysome loading or ribosome abundance. We have also taken a modeling approach to ask whether these changes are necessary to balance C consumption in protein synthesis with diurnal changes in the C supply. Overall polysome loading in wild-type Col-0 ranges between 65-70% in the light period, 40% at the end of the night and about 20% when starch is exhausted after a short extension of the night (Figs. 3-4, Supplemental Fig. S4). It falls to <25% during the night in the starchless pgm mutant. Values of 20-25% resemble those seen under a range of stress treatments, including including dehydration (Hsiao et al., 1970;Scott et al., 1979;Kawaguchi et al, 2003;2004;2005, Matsuura et al., 2010, anaerobiosis (Branco-Price et al., 2005;2008;Mustroph et al., 2009) and severe C depletion (Nicolai et al., 2006). We conclude that there are substantial changes in overall polysome loading during an undisturbed diurnal cycle, and that exhaustion of starch leads to a decrease comparable to that seen under extreme stress.

Compartment-specific changes in polysome loading
Protein synthesis occurs in three different subcellular compartments in plant cells; the cytosol, the plastid and the mitochondria. Chloroplast-encoded proteins like RBCL represent substantial proportion of total leaf protein; correspondingly the plastid accounts for a substantial proportion of the total ribosomes in photosynthetic cells (Fig. 8;Detchon and Possingham, 1972;Dean and Leach, 1982).
To assess the compartment-specific response of polysome loading we investigated the distribution of cytosolic, plastidic and mitochondrial rRNA species in polysome density gradients. This was done at selected times during the diurnal cycle (Fig. 5) and in an additional experiment in which leaves were illuminated in sub-compensation point CO 2 or ambient CO 2 to separate the impact of light-driven CO 2 fixation from the effect of light per se and (Fig. 6). Cytosolic polysome loading remained relatively high for most of the night, and the increase after illumination in the morning was dependent on provision of CO 2 to allow C-fixation. Mitochondrial polysome loading remained high throughout the day and night, and was unaffected by the CO 2 concentration. Plastidic polysome loading was strongly light dependent; it was high in the light, low in the night and increased in the light in sub-compensation point CO 2 . The latter is in agreement with many earlier studies showing that translation is strongly light dependent in plastids (Marin-Navarro et al., 2007).
The subcellular responses of polysome loading resemble the responses of the cytosolic, chloroplastic and mitochondrial phosphorylation potential to illumination. Studies in protoplasts from leaves of various species have shown that the cytosolic ATP/ADP ratio is high in the light and the dark, and that the mitochondrial ATP/ADP ratio remains unaltered or even increases slightly in the dark. In contrast, the plastidic ATP/ADP ratio is very low in the dark and increases in the light (Stitt et al., 1982;1983;Gardeström and Wigge, 1988;Igamberdiev et al., 2001). Non-aqueous fractionation of leaves has also shown that the plastidic ATP/ADP ratio is very low in the dark and increases in the light (Keys and Whittingham 1969;Dietz andHeber, 1984, Sellami 1976;Heineke et al., 1991). The latter technique is unable to separate the cytosolic and mitochondria (Supplemental Fig. S5) points to the majority of the plastidic polysomes containing only a small number of ribosomes in the dark. This is inconsistent with a general inhibition of elongation but is consistent with ribosome arrest at a small number of specific sites. The LPS/SPS ratio was also always low for mitochondrial polysomes. This is consistent with the predominance of short reading frames and the importance of translation regulation in the mitochondria (MacKenzie andMcIntosh 1999, Geigé et al. 2000) The differing diurnal response of polysome loading in the plastid and cytosol raises questions with respect to the coordination of translation in these two compartments.
Almost all of the plastid-encoded proteins are components of RuBisCO or large protein complexes in the thylakoid, which also contain nuclear-encoded proteins (Marin-Navarro et al., 2007). To avoid a cycle of synthesis and degradation of the nuclear encoded components in the dark, it appears necessary that their transcripts are rapidly degraded in the dark and/or that their translation is strongly decreased in the dark. Incidentally, any preferential inhibition of translation of the nuclear-encoded components of RuBisCO and thylakoid complex proteins would make more cytosolic ribosomes available for translation of other proteins in the dark.

Cytosolic polysome loading tracks sucrose content
Overall polysome loading closely tracks sucrose levels during diurnal cycles (Figs 1A, F and Fig. 3aA,B). Polysome loading changes rapidly in response to transient changes in sucrose levels when Col-0 rosettes are darkened (Fig. 3) and after adding sucrose to Cstarved seedlings (Supplemental Fig. S3). The transient decrease of sucrose after darkening is probably due to a delay until starch degradation commences (Fig. 2). Plants are seldom exposed to sudden darkening in their natural environment, and may not have evolved regulatory mechanisms that act to immediately activate starch degradation after a sudden transition. Although sudden darkening is an artificial treatment, it is a useful perturbation to uncover the close temporal connection between changes in sucrose and polysome loading.
A meta-analysis revealed a robust correlation between overall polysome loading and rosette sucrose content across a large set of experiments in wild-type Col-0 and pgm in diurnal cycles and in Col-0 in low CO 2 ( Fig. 4; Supplemental Fig. S4). The only data points that show a major deviation are at early times in the light period in pgm, when sucrose accumulates to very high levels, but polysome loading is still rising. Whilst a definitive explanation for the latter discrepancy is not possible, two explanations appear plausible. One is that the levels of sucrose in wild-type Col-0 may suffice to support maximal stimulation of polysome loading. The second is that, like root extension growth (Yazdanbakhsh et al., 2011), there may be a time lag before protein synthesis can be fully reactivated in pgm following a period of acute C starvation in the preceding night.
Compartment-resolved analyses  indicate that this correlation between polysome loading and sucrose is probably driven by changes in cytosolic polysome loading (Fig. 7).
Polysome loading correlated more strongly with sucrose than with other metabolites, in particular reducing sugars ( Fig. 4B; Supplemental Fig. S2). This is striking because reducing sugars are present at similar levels to sucrose, and might appear a more immediately readily-metabolised form of carbohydrate. However, sucrose is the form in which C is transported in plants. Whereas sucrose levels responded rapidly to changes in photosynthesis and starch breakdown, reducing sugars changed more slowly (Fig. 1).

Ribosome abundance does not change substantially during diurnal cycles
There is mounting evidence that growth is impaired by mutations in ribosome assembly  To investigate if these diurnal changes of transcripts result in changes in ribosome abundance we measured the absolute abundance of cytosolic, plastid and mitochondrial 18/16S rRNA as a proxy for ribosome number (Fig. 8). There were no significant changes in the abundance of cytosolic, plastid or mitochondrial rRNA in the diurnal cycle or a short extended night. However, there was a slightly lower abundance of plastid rRNA and an almost 2-fold increase in abundance of mitochondrial rRNA in pgm compared to wild-type Col-0. This indicates that whilst changes in C metabolism have little immediate impact on ribosome abundance, they do result in long term adjustments.
The increased level of mitochondrial ribosomes in pgm might speculatively be related to the high rates of respiration in this mutant at the start of the night (Caspar et al., 1985;Gibon et al., 2004a).
In microbes, excess ribosomes are degraded when polysome loading is low (Davis et al., 1986;Kuroda et al., 2001;Zundel et al., 2009). This does not happen at night in plants, even in the pgm mutant where polysome loading decreases strongly at night. Degradation of ribosomes during the night would necessitate their re-synthesis at the start of the next light period, which would represent a considerable energy load (Warner et al., 1999;Snoep et al., 2006;Houseley and Tollervey, 2009;Zundel et al., 2009) predicted that the rate of synthesis is of the same order as the rate of growth. It is likely that a similar situation holds for ribosomes. The turnover times of ribosomes in plants is not known, but is of the order of three days in mammalian liver (Hirsch and Hiatt 1966;Nikolov et al., 1987).

Modeling the balance between C availability and C consumption for protein synthesis at night
We used our quantitative data for ribosome abundance and polysome loading to model the rate of protein synthesis and the associated energy costs throughout the diurnal cycle ( Fig. 9). Comparison with the measured rates of starch degradation and respiration predicted that a decrease in protein synthesis is required at night to balance energy consumption with availability of C from starch degradation. These calculations also predicted that costs at night are decreased by using amino acids that are accumulated in the preceding light period (Supplemental Fig. S10).
These calculations require assumptions, in particular, that the rate of elongation is the same in the light and the dark and that all the ribosomes in the polysome fraction are involved in protein synthesis. We tested these assumptions by using 13 CO 2 labelling to obtain a qualitative estimate of the rates of protein synthesis in the light period and the dark. These measurements indicated that the changes in polysome loading may underestimate the decrease in protein synthesis at night. There are several explanations for this discrepancy. First, the rate of elongation might be decreased in the dark. Second, the ribosome distribution in polysome density gradients may overestimate active ribosomes because some of the ribosomes are arrested. Third, differences in 13 C enrichment in the free amino acid precursors may affect our estimates of the rate of protein synthesis. Potential sources of error include decrease in enrichment of free amino acids in the dark due to mobilisation of weakly labelled C reserves, or recycling of unlabelled amino acids released by degradation of unlabelled proteins. These are unlikely to be responsible for the large decrease in 13 C incorporation into protein in the dark. Our experimental protocol will ensure that starch and other major C reserves are almost completely labelled at dusk (see Results), and ten different amino acids provided similar estimates for the rate of protein synthesis in the night compared to the light (0.24-0.37), irrespective of the likelihood that the amino acid will be recycled before it is reequilibrated with labelled pools in central metabolism. The only outlier was Glu (0.87).
The C precursor for Glu is 2-oxoglutarate. It was recently shown that 2-oxoglutarate is synthesized from a preformed unlabelled pool of citrate in the light, and this citrate pool is replenished at night using C that was fixed in the previous light period (Tcherkez et al., 2012a;2012b;Szecowka et al., 2013). This may explain the outlier value for Glu in our experiments. In the future it will be desirable to obtain information about enrichment in free amino acid pools to further strengthen the estimates of flux to protein.
The overall distribution of ribosomes, and in particular cytosolic polysomes, in density is consistent with protein synthesis being mainly regulated by the rate of initiation  2006;Usadel et al., 2008), translation of some transcripts is specifically increased in Cstarvation (Nicolai et al., 2006) and specific proteins increase in C starved Arabidopsis rosettes (Gibon et al., 2004b;2006). C-starvation induces autophagy (Brouquisse et al., 1991;Aubert et al., 1996;Contento et al., 2004) . In agreement, metabolites that are released by catabolism of protein, cell wall and lipids increase in an extended night in Col-0, and during the night in pgm (Gibon et al., 2004a;2006;Usadel et al., 2008). They could provide energy to support a low rate of protein synthesis when starch is exhausted.

Optimisation of ribosome utilisation
The indirect costs of protein synthesis include the C, N and P invested in ribosomes and the costs of synthesizing and maintaining ribosomes (Warner, 1999;Rudla and Warner 2004;Snoep et al., 2006). Cells require high concentrations of ribosomes because the rate of ribosome progression along the mRNA is relatively slow (reviewed in Mathews et al., 2007) with typical values of four-five and seven-eight amino acids added per sec in animal cells at 25-26°C (e.g., Lodish and Jacobsen, 1972;Palmiter, 1974) and yeast cells at 30°C (Arava et al, 2003), respectively. The rate of progression is constrained by the size of the ribosome, the need to unwind secondary structures in the mRNA (Wen et al., 2008) and the need to proof read, which involves pausing after recruitment of an aminoacyl-tRNA to allow competition with other aminoacyl-tRNA species (Kramer and Farabaugh, 2007;Zaher and Green, 2009a;2009b).
Indirect costs would be minimized in Arabidopsis by maintaining high rates of protein synthesis through the entire 24 h cycle. However, as already discussed, protein synthesis is decreased at night to balance the direct costs with the rate of C release from starch. A decreased rate of protein synthesis at night will also reduce the direct costs incurred per 24 h cycle. The direct costs include (i) the ATP that is required for amino acid activation and peptide bond synthesis and (ii) the NAD(P)H, ATP and C that are required to convert nitrate into amino acids. In the light, ATP, NAD(P)H and C are provided by photosynthesis, whereas at night they are provided by catabolism of reserves like starch, with a concomitant loss of free energy (Penning de Vries, 1975;Hachiya et al, 2007;Amthor et al., 2010;Raven 2012).
To produce a given amount of protein per 24 h cycle, a decreased rate of protein synthesis at night will have to be counterbalanced by an increased rate of protein synthesis in the light. Given that ribosomes are already used intensively in the light, this would require an increase in ribosome abundance and, hence, increased indirect costs. Immediately following harvest, leaf tissue and seedlings were frozen in liquid nitrogen.
Samples were ground to a fine powder and sub-aliquoted at -70ºC using a cryogenic grinding robot (http://www.labman.co.uk/MAPC_Cryogenic_Grinder.html, Labman Automation Ltd., Stokesley, North Yorkshire, UK) and stored until analysis at stored at -80°C.

Polysome isolation and analysis
Polysomes were fractionated from crude leaf extracts as described previously  Supplemental Table S1.

Metabolites
Sucrose, glucose, fructose were determined in ethanol extracts as in Geigenberger et al., (1996), starch and glucose-6P as in Gibon et al., (2004a). and total amino acids as in     The three data points with the highest sucrose content are for pgm at the times 0.5, 1 and 2 h after illumination (see Fig. 1F).
The lower sub-panel shows the relation for log-transformed data. Plots of the relation between polysome loading and glucose, fructose and Glc6P are provided in Supplemental    Supplemental Table SIV. Color-coding is as in Fig. 5 The results are the mean ± SD of three independent biological replicates.  The original data are given in Supplemental Table SII. The results are the mean ± SD (n = 3 biological replicates).

Supplemental Material
Supplemental Figure S1: Representative example of polysome profiles at the end of night and 2 h into the light period.
Supplemental Figure S2. Regression plots of metabolites against polysome loading. The original data is given in Supplemental Table SII.
Supplemental Figure S3. Representative profile of polysome loading after addition of sucrose to C-starved Arabidopsis Col-0 seedlings. Supplemental Figure S7.      . The three data points with the highest sucrose content are for pgm at the times 0.5, 1 and 2 h after illumination. The lower sub-panel shows the relation for log-transformed data. Plots of the relation between polysome loading and glucose, fructose and Glc6P are provided in Supplemental Fig. S2. (B) Partial correlation analysis. The R-values are shown in the lower right hand sector and p-values (color coded for significance) in the upper right-hand sector. Partial correlation analysis was performed to exclude secondary correlations. The analysis was performed using log-transformed data; similar results were obtained with nonlogged data (p < 0.0005 for sucrose vs (SPS + LPS), and >0.05 for all other traits; not shown). The data for both panels are taken from Figs. 1 and 4A A Figure 5. Loading of cytosolic, plastidic and mitochondrial ribosomes into polysomes. Arabidopsis Col-0 WT was grown in a 12 h light / 12 h dark cycle as in Fig. 1, rosettes harvested after 12 h light (L) and 0.5, 1, 8 and 12 h darkness (D) and subjected to polysome gradient fractionation. To determine the abundance of cytosolic, plastidic and mitochondrial ribosomes in each gradient fraction, cDNA was prepared after adding eight external standands and rRNA quantified by qRT-PCR using specific primer pairs for the cytosolic 18S, the plastidic 16S and mitochondrial 18S rRNA species. (A) Comparison of polysome loading as calculated from (1) Table SIV. Color-coding is as in Fig. 5. The results are the mean ± SD of three independent biological replicates.