Variation in leaf dark respiration among C₃ and C₄ grasses is associated with use of different substrates Open Access

Abstract

Measurements of respiratory properties have often been made at a single time point either during daytime using dark-adapted leaves or during nighttime. The influence of the day–night cycle on respiratory metabolism has received less attention but is crucial to understand photosynthesis and photorespiration. Here, we examined how CO₂- and O₂-based rates of leaf dark respiration (R_dark) differed between midday (after 30-min dark adaptation) and midnight in 8 C₃ and C₄ grasses. We used these data to calculate the respiratory quotient (RQ; ratio of CO₂ release to O₂ uptake), and assessed relationships between R_dark and leaf metabolome. R_dark was higher at midday than midnight, especially in C₄ species. The day–night difference in R_dark was more evident when expressed on a CO₂ than O₂ basis, with the RQ being higher at midday than midnight in all species, except in rice (Oryza sativa). Metabolomic analyses showed little correlation of R_dark or RQ with leaf carbohydrates (sucrose, glucose, fructose, or starch) but strong multivariate relationships with other metabolites. The results suggest that rates of R_dark and differences in RQ were determined by several concurrent CO₂-producing and O₂-consuming metabolic pathways, not only the tricarboxylic acid cycle (organic acids utilization) but also the pentose phosphate pathway, galactose metabolism, and secondary metabolism. As such, R_dark was time-, type- (C₃/C₄) and species-dependent, due to the use of different substrates.

Introduction

Increasing attention is being given to how rates of leaf respiration measured in the dark (R_dark) differ among species and environments (Wright et al. 2006; Atkin et al. 2015). However, R_dark data used in these studies were collected during the day, following 30 min of dark exposure to avoid the postillumination photorespiratory CO₂ burst and light-enhanced dark respiration (LEDR), 2 common postillumination transients (Azcón-Bieto et al. 1983; Azcón-Bieto and Osmond 1983; Reddy et al. 1991; Xue et al. 1996; Atkin et al. 1997; Atkin et al. 1998). After 30-min of dark exposure, it is often assumed that R_dark and its associated metabolism revert to a nocturnal phenotype (Atkin et al. 2000; Padmasree et al. 2002). Such an assumption allows for measurements of nighttime R_dark to be modeled using daytime R_dark (Atkin et al. 2008; Huntingford et al. 2017; Butler et al. 2021). However, there is some evidence that even after 30 min of dark exposure during the day, metabolite and transcript profiles remain influenced by conditions in the preceding photoperiod (Noguchi et al. 2005; Florez-Sarasa et al. 2009), potentially affecting rates of R_dark through differences in substrate availability and demands for respiratory products (but see Florez-Sarasa et al. 2012). For example, isotopic and enzymatic analyses of castor bean leaves (Ricinus communis) sampled during a diel cycle suggest that dark-adapted respiratory metabolism differs between the light and the dark phase (Gessler et al. 2009).

In most studies, R_dark is measured as either CO₂ efflux or O₂ uptake, but rarely both (Wright et al. 2006; Atkin et al. 2015; Scafaro et al. 2017). This is not an issue when the respiratory quotient (RQ; the ratio of respiratory CO₂ release to O₂ uptake) is at unity (i.e. when soluble sugars are the only substrate for R_dark; Lambers et al. 2008), with the rate of R_dark being the same irrespective of whether the measurements are CO₂- or O₂-based. However, if at any given time during the day or night, respiration utilizes organic acids rather than sugars as substrates, then the rate of respiratory CO₂ release per unit O₂ uptake (and per unit ATP produced) will be higher (i.e. RQ > 1.0) (Lambers et al. 2008). It has been shown that C₃ wheat leaves (Triticum aestivum) exhibit RQ value of 1.8 just after a period of illumination, although such value dropped to 0.9 overnight (Azcón-Bieto et al. 1983). In French bean leaves (Phaseolus vulgaris), RQ changes within the first few hours of night and with temperature (Tcherkez et al. 2003). Similarly, wheat grown during warm nights had a greater acclimation-induced reduction in O₂- than CO₂-based R_dark, pointing toward adjustments of RQ in warm-acclimated plants (Coast et al. 2021; Posch et al. 2022). Environmental influences on RQ are important because almost all terrestrial biosphere models predict CO₂-based rates of R_dark based on the link between leaf nitrogen (N) and demands for ATP, with ATP being required to support turnover of leaf proteins such as Rubisco (Atkin et al. 2017). However, if leaves use organic acids rather than sugars as a respiratory substrate, there will be relatively more respiratory CO₂ released per O₂ taken up and ATP molecule produced, with O₂-based R_dark being a better indicator of ATP production than its CO₂-based R_dark counterparts in cases where the RQ is greater than unity (e.g. when organic acids fuel respiration). Organic acids (particularly malate) are known to accumulate during the day and decrease at night in C₃ leaves (Urbanczyk-Wochniak et al. 2005; Zell et al. 2010; Rashid et al. 2020). By contrast, respiratory substrates (especially organic acids) are usually more abundant in C₄ leaves than C₃, because of the CO₂-concentration mechanism. For example, stable isotope studies reported that metabolite pools of malate, pyruvate and 2-oxoglutarate were 2-fold higher in C₄ NADP-ME type than C₃ species (Arrivault et al. 2017; Borghi et al. 2022). These organic acids also remained high throughout the night in a range of C₄ species, compared to C₃ plants (Hatch 1979; Stitt and Heldt 1985; Du et al. 2000; Czedik-Eysenberg et al. 2016). Hence, it is possible that R_dark and RQ of C₄ plants differ from C₃ plants and between the day and night, because of day–night shifts in the availability and utilization of organic acids.

R_dark and RQ values can also be altered by sugar metabolism. C₃ and C₄ leaves exhibit differences in the regulation of nonstructural carbohydrates (i.e. starch and soluble sugars), which, in turn, may lead to contrasting availability of respiratory products. C₃ and C₄ leaves accumulate transitory starch during the day, and degrade starch to sucrose at night (Fünfgeld et al. 2022). Sucrose could be used locally as a substrate by respiration or exported to other tissues for growth and maintenance (Atkin et al. 2000). Although the fates of sucrose and starch may be similar in C₃ and C₄ plants, the rate of starch and sucrose synthesis during the day, and rate of sucrose export in the day and night likely differ. There is evidence suggesting that C₄ plants have higher potential for soluble sugar and starch production during the day due to higher photosynthesis, compared to C₃ (Weise et al. 2011). In addition, both Grodzinski et al. (1998) and Bouma et al. (1995) reported higher nocturnal sucrose export rates (and higher overall rates across a 24-h period), in a range of C₄ monocots and eudicots, compared to most of their C₃ counterparts. Given that sucrose export (i.e. phloem loading) requires ATP (Bouma et al. 1995), it could impact on R_dark. However, to our knowledge, it has not been examined how day–night changes in sucrose levels may correspond with variation in R_dark and be reflected in RQ in C₃ and C₄ leaves.

Here, we investigated whether R_dark varies during the day–night cycle and how it relates to potential respiratory substrates in 8 C₃ and C₄ grasses. We examined how rates of R_dark measured at midday (following 30 min of exposure to darkness) differed from those measured at midnight (following 6 h of darkness), and whether such differences vary when measured on a CO₂ or O₂ basis. Leaf soluble sugar and starch contents were determined and metabolomic analysis was performed. Potential relationships between R_dark (or RQ) and metabolites were examined by univariate and multivariate statistics. We addressed the following questions: (i) How do respiratory substrates (soluble sugars and organic acids) vary between day and night in dark-adapted leaves and do they correlate with variations in rates of R_dark or RQ values? (ii) Are there any differences in the leaf metabolome between C₃ and C₄ species and do they correlate to changes in R_dark or RQ values? (iii) What are the metabolic drivers of day–night variations in R_dark or RQ values of dark-exposed leaves in C₄ plants?

Results

R_dark decreased from midday to midnight

When measured after 30 min of dark exposure, CO₂-based R_dark was higher at midday than when measured at midnight after 6 h of darkness (P < 0.001; Fig. 1A). CO₂-based R_dark of all examined species were significantly faster at midday than midnight, except for Astrebla lappacea (P < 0.001; Fig. 1A). The difference between midday and midnight values was less pronounced when R_dark values were measured on a O₂ basis, with rates being significantly greater at midday compared to midnight only in C₄ species sorghum (Sorghum bicolor), Setaria viridis, and Panicum coloratum (P < 0.01; Fig. 1B). Rates of O₂-based R_dark differed significantly among C₃ and C₄ species at each timepoint (P < 0.001), with the differences among species being greater at midday than midnight (i.e. a significant time $×$ species interaction, P < 0.05; Fig. 1B).

Total nonstructural carbohydrates did not change with time

When measured after dark exposure, soluble sugar levels differed significantly among species at a given time (Fig. 2A to C; P < 0.001). For example, the C₄ NAD-ME type P. coloratum and A. lappacea exhibited significantly higher concentrations of glucose than the rest of the species at midday (P < 0.001; Fig. 2A), and higher fructose concentrations at both midday and midnight (P < 0.001; Fig. 2B). There was a significant interaction effect between time and species for glucose and sucrose concentration, mostly driven by P. coloratum and A. lappacea (Fig. 2A and C). By contrast, there was no interaction between time and species for total soluble sugar concentration, which remained rather stable through time (Fig. 2D). Thus, while individual sugars varied between midday and midnight in a species-dependent manner, all species showed similar total soluble sugar concentration at the 2 sampling timepoints. The starch content differed significantly amongst species, being higher in C₄ compared to C₃ species (P < 0.001; Fig. 2E). There was also a significantly higher starch content at midnight compared to midday, with the increase being mostly due to the change in Panicum maximum (P < 0.05; Fig. 2E). That is, midday starch content was considerably lower than that at midnight in P. maximum, perhaps because of a low rate of nocturnal starch degradation (see Discussion). Overall, total nonstructural carbohydrate (TNC, soluble sugars and starch) content was rather stable in most C₃ and C₄ species, except for rice (Oryza sativa; C₃) and P. maximum (C₄ PCK type) (P = 0.01 and 0.03, respectively; Fig. 2F). At each timepoint, the 2 C₄ NAD-ME type species and NADP-ME type sorghum showed significantly higher TNC levels than C₃ and the rest of C₄ species (P < 0.001; Fig. 2F), revealing minimal differences between photosynthetic types (C₃ vs. C₄).

Leaf metabolomes were influenced more by photosynthetic types than time

Gas chromatography-mass spectrometry (GC-MS) metabolomics quantified 47 major metabolites in examined C₃ and C₄ species (Supplementary Fig. S1 and Data Set 1). We analyzed the data using a 2-way ANOVA, taking photosynthetic type and sampling timepoint as factors. Thirty metabolites were influenced by photosynthetic types, while only 5 were affected by sampling timepoint (Fig. 3A and B). Metabolites that were affected by photosynthetic type formed 3 main clusters (Fig. 3A; red frames). The first cluster mostly consisted of organic acids (e.g. citrate or malate) that were more abundant in C₄ NAD-ME and NADP-ME species. The second cluster included glycerate, glycerol 3-phosphate, quinate and shikimate and was mostly represented in C₄ PCK types. The 3rd cluster was made of amino acids and sugar derivatives (e.g. myoinositol) in C₃ species. Metabolites that were influenced by the sampling timepoint were (iso)leucine (more abundant during the night), serine, glycerate and succinate (more abundant during the day) (Fig. 3B). The interaction effect (type $×$ timepoint) was seen in 4 metabolites (Fig. 3C). Glycerate and glycerol 3-phosphate were more abundant during the day in C₄ species, while N-acetylglutamate and serine were more abundant in C₃ species during the day (Fig. 3C).

Strong clustering (covariation) was observed for glycerate and glycerol 3-phosphate in C₄ species, with such pattern being associated with an increase in quinate and shikimate abundance (Fig. 3A). This suggests that the clustering of glycerate and glycerol 3-phosphate is related to secondary metabolism (Fig. 3D), rather than primary metabolism (i.e. photorespiration or glycolysis) where these metabolites also play a role. Such an interpretation agrees with our understanding of C₄ plants that have minimal photorespiratory activity. Many C₄ grasses have a specific secondary metabolism leading to the production of benzoxazinoids, which involves the aforementioned metabolites (Fig. 3D) (Frey et al. 2009). Strong daytime clustering of N-acetylglutamate and aspartate contents in C₃ species indicates that benzoxazinoids metabolism is more pronounced during the day (Fig. 3A and E). In C₃ species, higher contents in serine and N-acetylglutamate are likely reflective of photorespiration and ammonium recycling through polyamine metabolism (Fig. 3E) (Blume et al. 2019; Timm et al. 2021). This is consistent with our finding that aspartate was the closest covariant metabolite with N-acetylglutamate (Fig. 3A).

Since there was little overall day–night difference in metabolites, we examined time-specific responses of metabolites in individual species using principal component analysis (PCA) (Supplementary Fig. S2). There was a separation of midday and midnight samples in all species, except in C₄ PCK type Urochloa panicoides (Supplementary Fig. S2H). Interestingly, the distribution of dark-exposed samples was not driven by a single metabolite class, although in C₄ NADP-ME type sorghum midnight samples were partly driven by soluble sugars (Supplementary Fig. S2E). In summary, there was no consistent time-specific metabolome difference across species, and the C₃/C₄-specific difference in metabolites may be attributed to multiple pathways.

Tricarboxylic acid pathway intermediates showed minimal time effect

We next examined changes in tricarboxylic acid pathway (TCAP) intermediates and their derivatives to gain insight into how variations in metabolites might be linked to respiratory metabolism (Fig. 4). Similar to findings presented in Fig. 3A, there was a significant species effect on all TCAP organic acids (Fig. 4A), with succinate being affected by both time and photosynthetic type. Given the time effect on succinate was observed in both C₃ and C₄ species (Fig. 4A) and C₄ plants have minimal photorespiration, changes in succinate over time were likely not a result of photorespiration (via the γ-aminobutyrate, GABA, shunt) but possibly the biosynthesis of oxalate via isocitrate lyase (Fig. 4B). We also found that malate content was not affected by C₃/C₄ photosynthetic type. Knowing that malate plays a crucial role in C₄ photosynthesis but not in C₃, our result suggests that 30 min of dark exposure was sufficient to minimize the effect of C₄ photosynthesis on TCAP intermediates. Taken together, there seems to be no systematic day–night-specific difference in TCAP metabolism of dark-adapted C₃ and C₄ leaves.

Apparent RQ reflects changes in substrate usage

Given the contrasting day–night patterns of CO₂ and O₂ exchange (Fig. 1), we calculated “apparent” RQ values by dividing CO₂- by O₂-based R_dark at midday and midnight (Table 1). Here, the term “apparent” is used because CO₂ evolution and O₂ consumption were measured on distinct samples with different methods. Due to this limitation, we will focus on relative changes in RQ, but not on absolute RQ values. Apparent RQ values were greater at midday than midnight in all species, except in C₃ rice (Table 1). In C₃ barley (Hordeum vulgare), C₄ sorghum and P. maximum, the apparent RQ values declined from midday to midnight, suggesting a shift in respiratory substrate utilization to relatively less oxygenated substrates at night (e.g. a decrease in the organic acid-to-sugar utilization ratio with time). By contrast, in rice, the apparent RQ value remained stable at unity, suggesting that the utilization of respiratory substrates (perhaps soluble sugars) did not change with time (Table 1).

Patterns of R_dark and RQ were codriven by multiple metabolites

To investigate the origin of metabolic pathways responsible for respiratory O₂ consumption, we examined quantitative relationships between metabolite contents and R_dark (and apparent RQ) using multivariate analysis. Since our dataset consists of a large number of samples (90) compared to the number of major metabolites quantified (47), the reduction of dimensionality and the identification of major drivers of respiration via orthogonal projection on latent structure (OPLS) analysis are expected to be robust. We first took O₂-based R_dark as a Y quantitative response variable, given that this trait was measured on the same leaf sample as the metabolome (Fig. 5). The OPLS model discriminated samples along the x axis with respect to R_dark (Fig. 5A), with a good explicative power (R² = 0.61, Q² = 0.33) and high statistical significance (P_CV-ANOVA = 5.4 $×$ 10⁻⁶). As such, there was a good relationship between observed and OPLS-predicted R_dark (Fig. 5B). Only a few samples belonging to C₃ rice were outside the Hotelling's ellipse (Fig. 5A). Sample discrimination was not driven by time (Supplementary Fig. S3). Sugars (ribose, lactose, fructose), ascorbate, sugar and ascorbate derivatives (galactinol, threitol, myoinositol, ribonate) and some amino acids (serine, glycine, aspartate) were positively correlated with R_dark, while citrate, isocitrate, and γ-aminobutyrate were negatively correlated with R_dark (Fig. 5C).

An OPLS analysis was also conducted to explore relationships between apparent RQ and leaf metabolome. Since CO₂ evolution was not measured on the same leaf sample that used in metabolomics, we adopted a randomized approach. That is, we assigned a random CO₂-based R_dark value to a leaf sample and used this R_dark to calculate apparent RQ and compute the OPLS model. This process was reiterated with different assignments of CO₂-based R_dark (i.e. permutation) to confirm the robustness of OPLS analysis. Sample discrimination with respect to apparent RQ values was seen along the x axis (0.43 < R² < 0.58, 0.15 < Q² < 0.34 and 2 $×$ 10⁻⁴ < P_CV-ANOVA < 0.05; Fig. 6A). The effect of swapping CO₂-based R_dark (i.e. differences between multiple OPLS models) was found to be modest (Fig. 6B). The most correlated metabolites with apparent RQ appeared to be organic acids (glycerate, succinate, and malate, positively related), sugar phosphates and myoinositol (negatively related) (Fig. 6B). Interestingly, when examining the relationship between apparent RQ values (calculated using averaged CO₂-based R_dark) and metabolites, positive correlations were found between midnight apparent RQ values and malate (R² = 0.79 and P < 0.01; Supplementary Fig. S4A) and aspartate concentration ratios (R² = 0.55 and P = 0.03; Supplementary Fig. S4B). However, it should be noted that the proportion of variance explained by individual metabolites was low (up to 2.5% only; Fig. 6C). This indicates that apparent RQ was not explained by a single metabolite, but by several concurrent metabolic pathways.

Discussion

Lack of responses of carbohydrates to time

We found a lack of changes in soluble sugar contents at midday (measured after a 30-min dark adaptation) compared to that at midnight among all C₃ and C₄ species (Fig. 2). This result agrees with a handful of studies that showed the levels of soluble sugars (particularly sucrose) remain stable over a day–night cycle in C₃ and C₄ source leaves (Sicher et al. 1984; Bläsing et al. 2005; Gebauer et al. 2017; De Souza et al. 2018; Gersony et al. 2020; Rashid et al. 2020; Wang et al. 2020). However, a number of studies reported the opposite and showed that soluble sugar contents are higher in leaves harvested during the photoperiod than at night (e.g. Azcón-Bieto and Osmond 1983; Noguchi et al. 1996; Noguchi and Terashima 1997; Florez-Sarasa et al. 2012). We suggest that such differences may be species or growth form specific (e.g. monocots versus eudicots), and a study comparing sugar regulation among species of different plant functional types would be informative.

Interestingly, our result also showed similar starch content at midday and midnight in all species except C₄P. maximum (see Results and Fig. 2E). This seems inconsistent with starch circadian regulation, where starch accumulates during the day and degrades overnight (see Graf and Smith 2011 for a review). We however recognize that our study only provides a snapshot of the starch content at midday and midnight, and does not infer the rate/pattern of starch accumulation/degradation. When starch accumulates and degrades at a similar rate over a 12 h diurnal cycle, the starch content at midday and midnight would be comparable. The starch content of C₄P. maximum at midnight is significantly higher than that of midday, suggesting a slower nocturnal starch degradation (Fig. 2E). Further study is needed to map out changes in soluble sugar/starch contents with higher temporal resolution.

Response of R_dark and metabolite profiles to time

Our results show that the average rates of CO₂-based R_dark measured at midday were generally higher than those measured at midnight (Fig. 1). Since leaves were dark-exposed for 30 min, it is unlikely that this effect came from the postillumination photorespiratory CO₂ burst. We recognize that higher R_dark at midday (relative to midnight) may to some extent be due to LEDR (Barbour et al. 2007; Gessler et al. 2009). Higher R_dark during LEDR is related to malate decarboxylation, as suggested by the rapid decline in both the malate content and the natural carbon isotope composition (δ¹³C) of evolved CO₂ (Gessler et al. 2008, 2009). Our results nevertheless highlight that there was no consistent day–night change in malate content across species (Fig. 4), showing that malate utilization was mostly related to species-driven differences in R_dark (further discussion about malate utilization is provided in Supplementary Note 1). By contrast, the difference between O₂-based R_dark measured at midday and midnight was less pronounced, with 5 out of the 8 examined species (i.e. 2 C₃, 2 C₄ PCK types, and C₄ NAD-ME A. lappacea) showing similar R_dark at both timepoints (Fig. 1B). Our result in C₃ rice and barley agrees with previous experiments in Arabidopsis (Arabidopsis thaliana) (Florez-Sarasa et al. 2012).

We found that succinate was influenced by sampling timepoints (Fig. 3B) and correlated to apparent RQ (Fig. 6). In principle, succinate may contribute to CO₂ evolution via: (i) its synthesis from 2-oxoglutarate in the TCAP; (ii) the GABA shunt; and, (iii) its synthesis from isocitrate via oxalate (Fig. 4B; Supplementary Fig. S4). Since darkened leaves during daytime show a decline in 2-oxoglutarate dehydrogenase activity compared to darkness (Gessler et al. 2009), assumptions (ii) and (iii) are more likely. Additionally, day–night differences in soluble sugars (glucose, fructose, sucrose) were modest (Fig. 2), suggesting that they were unrelated to time-driven changes in R_dark. Further, apparent RQ values were negatively related to glucose 1-phosphate and mannose 6-phosphate (Fig. 6), indicating that nonrespiratory sugar metabolism may have affected CO₂ evolution and/or O₂ consumption (see below).

Our results also demonstrate that the contents in most metabolites measured after 30 min of dark adaptation were similar to those measured at midnight (Fig. 3B). This result differs from a report in Arabidopsis, where significant changes were observed in metabolite profiles quantified after 30 min of darkness compared to those at night (Florez-Sarasa et al. 2012). We recognize that this disagreement could be partially due to distinct metabolism in C₃ and C₄ species that leads to various responses of metabolome to time (Fig. 3A). For example, it has been shown that pool sizes of organic acids remained high through the night in C₄ species compared to their C₃ counterparts (Hatch 1979; Stitt and Heldt 1985; Du et al. 2000; Czedik-Eysenberg et al. 2016). Thus, it is possible that high levels of organic acids at night in C₄ species (relative to C₃) diminish the differences in metabolite profiles at midday and midnight in our study.

Metabolic correlation with R_dark and apparent RQ

Multivariate analysis showed that O₂-based R_dark was related to sugar species derived from the pentose phosphate pathway (ribose), galactose metabolism (galactinol, ascorbate, lactose), hexose phosphates interconversions (mannose 6-phosphate), organic acids from the TCAP (citrate, isocitrate) and other pathways (GABA, isothreonate, oxalate) (Fig. 5). In fact, several reactions in these metabolic pathways lead to O₂ consumption or production of NAD(P)H that must be reoxidized (thus consuming O₂) (Supplementary Fig. S5). For example, ascorbate can be synthesized from galactose through oxidation steps, with this process potentially influencing O₂-based R_dark. In addition, ascorbate degradation generates oxalate, which can be further oxidized to release CO₂ (Davies and Asker 1983; Green and Fry 2005; DeBolt et al. 2006). Interestingly, the most correlated metabolites with O₂-based R_dark were independent of time (Supplementary Fig. S3) and species (Fig. 5B), suggesting that these pathways appeared to be of importance when determining O₂-based R_dark of darkened leaves, regardless of species and sampling time. Overall, our results largely agree with a similar study in Arabidopsis, highlighting that changes in organic and amino acids (rather than conventional carbohydrates such as sucrose and fructose) were better correlated with R_dark (Florez-Sarasa et al. 2012), albeit the variation between C₃ and C₄ photosynthetic pathways could play a bigger role than the substrate types in determining R_dark (see below).

The multivariate analyses suggest that the apparent RQ values were driven by several pathways: sugar phosphates, galactose metabolism and specific organic acids (malate, succinate and glycerate) (Fig. 6B). In principle, sugar phosphates and galactose metabolism can influence the RQ via O₂ consumption (via oxidation and NAD(P)H reoxidation) and CO₂ evolution (via pentose phosphates and ascorbate metabolism) (Supplementary Fig. S5). While succinate metabolism is mostly associated with the day–night RQ difference (Figs. 3B and 4A), malate and glycerate are shown to contribute to species-driven RQ differences (Fig. 3A). Interestingly, glucose 1-phosphate, galactinol, lactose and myoinositol, that are related to either O₂-based R_dark or apparent RQ, are all intermediates of galactose metabolism, which in turn generates raffinose, an export form of sugar for phloem loading. The export of raffinose could potentially alter R_dark through ATP demands. Apoplastic export requires ATP and thus stimulates respiratory activity to energize sucrose transport (Bouma et al. 1995). Raffinose metabolism is related to symplastic loading, which also requires energy in the form of UTP for sugar interconversions (Hannah et al. 2006). Past studies have reported that sucrose is exported at higher rates during the day in C₄ unlike C₃ monocots (Grodzinski et al. 1998) but to our knowledge, there is no information on raffinose synthesis rate and how it compares between C₃ and C₄ species.

Respiratory differences between photosynthetic types

Overall, there was a higher CO₂-based R_dark in C₄ species at midday (Fig. 1A), contrasting to no consistent C₃/C₄ difference in O₂-based R_dark (Fig. 1B). C₄ NAD-ME species showed relatively higher apparent RQ values, compared to their C₃ counterparts (Figs. 6A and Supplementary Fig. S4). These results suggest that there were likely differences in respiratory substrates or the balance between CO₂ evolution and O₂ consumption due to specific metabolic pathways. In rice, where the apparent RQ value was not far from unity (Table 1), soluble sugars were likely consumed by respiratory metabolism. This is supported by published findings in rice (Noguchi et al. 2018) and Arabidopsis (Zell et al. 2010), and agrees with our multivariate analysis (Fig. 6B).

We also found a link between apparent RQ values and malate (and aspartate) content at midnight in C₄ NAD-ME and PCK type species (Supplementary Fig. S4), with malate being correlated with apparent RQ (Fig. 6B). The utilization of organic acids is expected to be more pronounced in C₄ species where mitochondrial enzymes are capable of processing C₄ acids. In C₄ NAD-ME and PCK types, mitochondria in bundle sheath cells play a direct role in photosynthetic CO₂-concentrating mechanism, and maximal activities of mitochondrion-located aspartate aminotransferase, NAD-malic enzyme and malate dehydrogenase have been found to be up to 6 times higher than in their C₃ and NADP-ME counterparts (Hatch et al. 1982, 1988; Hatch and Carnal 1992; Fan et al. 2022a). In addition, these mitochondria can process malate in the dark, while C₄ NADP-ME species lack the ability to utilize C₄ acids in darkness (Agostino et al. 1996; Fan et al. 2022b). We recognize that malate metabolism in C₃ species can take place at a low rate in darkness, and it is further discussed in Supplementary Note 1.

Differences in nonrespiratory metabolic pathways could also alter CO₂ production and/or O₂ consumption thus affecting R_dark in C₃ and C₄ leaves. For example, benzoxazinoids synthesis contributed to the metabolic differences in C₃ and C₄ species (Fig. 3). This pathway is a major metabolic route in grasses (Corcuera 1984; Gierl and Frey 2001; Frey et al. 2009), with known differences between taxonomic groups: benzoxazinoid metabolism is absent in barley and rice, but present in Paniceae species of the Urochloa/Echinochloa tribe (Wu et al. 2022). Importantly, benzoxazinoid metabolism can lead to CO₂ production and regenerate glycerol 3-phosphate that may be recycled to glycerate and phosphoenolpyruvate (Fig. 3D and Supplementary Fig. S5). This pathway also affects O₂ consumption through oxidation of NADH. In addition, ascorbate metabolism may differ between photosynthetic types, given that ascorbate content was significantly higher in C₃ leaves (Fig. 3) and correlated to O₂-based R_dark (Fig. 5).

Implications for modeling global patterns of leaf R_dark

The results of our study raise a couple of issues relevant to how leaf R_dark is presented in the land surface component of earth system models (ESMs). Current ESMs predict respiratory CO₂ release from C₃ leaves using an assumed relationship between R_dark and total leaf [N] and/or photosynthesis (i.e. maximum rate of Rubisco carboxylation; V_cmax) (Atkin et al. 2015; Huntingford et al. 2017; Butler et al. 2021). Such assumptions come from the fact that R_dark provides ATP needed to repair proteins, and the rates of protein turnover positively scale with the amount of proteins in leaves (i.e. leaf [N], particularly Rubisco content) (Atkin et al. 2017). What is not considered in the models, however, is how different substrates could potentially alter the rate of CO₂ release associated with a given rate of ATP production, and the efficiency of ATP production per unit of O₂ consumed. While CO₂ is released from metabolism (such as the TCAP), ATP is produced by the subsequent mETC coupling to respiratory O₂ uptake (Plaxton and Podestá 2006). This will not be an issue if soluble sugars were the major respiratory substrates, since the predicted CO₂ release would likely be a good estimation of the ATP production (i.e. RQ ≍ 1.0). However, if other metabolites are used, such as organic acids, as our results suggest in some C₃ and C₄ species, it is possible that the actual CO₂ cost per unit of ATP synthesized will be greater than previously thought. In addition, electron flow through the mETC involves both the cytochrome c oxidase (COX) and the alternative pathways. While both pathways complete for electrons and consume O₂, only the phosphorylating COX pathway generates ATP (Plaxton and Podestá 2006). It has been reported that the alternative pathway in leaves could mediate up to 60% of the total respiratory flux, thus reducing ATP production per unit of O₂ uptake (Ribas-Carbo et al. 2005). The issues of substrate types and mETC electron partitioning will have consequences on the prediction of nocturnal rates of R_dark using daytime measurements, and on modeling of R_dark from leaf [N] and V_cmax values.

A recent analysis of nocturnal R_dark of 31 C₃ species inhabiting in distinct biomes suggests that changes in leaf temperature through the night only account for less than one-half of the observed variation in nocturnal R_dark (Bruhn et al. 2022). The authors highlighted potential factors that may account for the nontemperature control of nocturnal R_dark, such as changes in the concentration of respiratory substrates, demand for respiratory products, relative engagement of alternative oxidase and changes in rates of other decarboxylation processes (Fondy and Geiger 1982; Hendrix and Huber 1986; Noguchi and Terashima 1997; Grimmer and Komor 1999; Matt et al. 2001; Svensson and Rasmusson 2001; Dutilleul et al. 2003; Bruhn 2023). A very similar overnight pattern in R_dark and concurrent changes in RQ and δ¹³C signature have been observed at the leaf scale (Tcherkez et al. 2003), in addition to an interaction with temperature. Overall, it suggests that changes in metabolic pathways are essential to explain nighttime variations in R_dark. Our study suggests that variation in nonrespiratory metabolic pathways (including secondary metabolism) can significantly contribute to changes in R_dark, as might day–night variations in certain substrates used to fuel R_dark. Further work is needed to understand the roles of day–night changes in substrate use and metabolic pathways in leaf R_dark measured during daytime and nighttime in a wide range of species, not just in C₄ grasses but also C₃ plant functional types represented in ESMs.

Limitations and future perspectives

One limitation of this study is the use of apparent RQ values and changes in metabolite pool sizes to infer respiratory substrates. As mentioned above and in Supplementary Note 1, apparent RQ values calculated using 2 instruments (i.e. gas-exchange analyzer and a fluorophore sensor) could lead to potential errors, albeit that the day–night relative change in apparent RQ values and the comparisons among species at a given timepoint should still be valid. Nevertheless, simultaneous measurement of respiratory CO₂ and O₂ fluxes would be crucial to determine RQ values accurately. Online membrane inlet MS (Beckmann et al. 2009), coupled CO₂/O₂ gas-exchange analyzer (Willms et al. 1997) and isotopic composition of naturally respired CO₂ (Tcherkez et al. 2003; Ghashghaie et al. 2016) are potential tools to quantify absolute RQ and to inform the class of the substrates (e.g. organic acids, carbohydrates or amino acids). To pinpoint the exact substrate used by R_dark, isotopic tracing mated with metabolome profiling would also be necessary (Jang et al. 2018; Florez-Sarasa et al. 2019).

Finally, identifying to what extent rates of R_dark vary among C₃ and C₄ species will be important. It has been reported that inter-specific rates of R_dark could vary 5-fold depending on plant size and nitrogen content (Reich et al. 2006). Rates of R_dark vary intra-specifically up to 2-fold as shown in Arabidopsis (O’Leary et al. 2017) and wheat (Scafaro et al. 2017). In this study, we have selected monocot species of closely related lineages, controlled the experimental environments and nutrient supply with a state-of-the-art growth facility, and measured leaves at similar developmental stages (i.e. the most recent fully expanded 4th leaves of the main tiller), in order to reduce developmental variation.

Our results show that R_dark varies between day (in darkened leaves) and night, and also between photosynthetic types, with no systematic difference being seen among C₃ and C₄ grasses. Metabolic determinants of R_dark or RQ appear to be: (i) related to multiple pathways including those involving classical respiratory substrates such as TCAP intermediates; or, (ii) dependent on whether the day/night or inter-species, carrying out carbon isotope labeling on all pathways of interest at this scale would be a challenge. R_dark differences were examined. Isotopic tracing would be helpful to pin down differences in CO₂ and O₂ exchange among these pathways. Yet, new technologies such as high-throughput isotope-assisted GC-MS analysis would be desired (Abadie et al. 2022). In many cases, leaves appear to be using organic acids to fuel R_dark during the day, with the result that daytime rates of CO₂-based R_dark are higher than those at night, even when O₂-based R_dark is relatively similar at midday and midnight. Underpinning these observations is evidence that malate and/or aspartate are likely used as respiratory substrates, particularly in C₄ NAD-ME and PCK type leaves, with possible consequences for the CO₂ cost of ATP synthesis—a finding that has implications for how R_dark is modeled in ESMs.

Materials and methods

Plant materials and harvesting

The species used in this study were: C₃ barley (Hordeum vulgare cv. Golden Promise) and rice (O. sativa cv. Takanari); C₄ NAD-ME type A. lappacea and P. coloratum; C₄ NADP-ME type sorghum (S. bicolor; cv. A66) and S. viridis (cv. A10); and, C₄ PCK type P. maximum and U. panicoides. Seeds were germinated and grown in 3-L pots in organic potting mix supplemented with slow-release fertilizer (Scotts Osmocote, Bella Vista, Australia). Pots were placed in a completely randomized order in controlled Climatron growth cabinets (Thermoline Inc.). Temperature was set to 30/25 °C day/night, with a 12-h photoperiod (photoperiod started at 7:00 Am), and photosynthetically photon flux density (PPFD) was ≍500 μ mol quanta m⁻² s⁻¹ at plant height.

The most recent fully expanded leaves (i.e. generally the 4th leaf of the main tiller) of 4-week-old plants were harvested 6 and 18 h after the beginning of the photoperiod, equivalent to 1:00 Pm (referred to as “midday” thereafter) and 1:00 Am on next day (referred to as “midnight” thereafter), respectively. Harvested leaves were then transported to the lab for 30-min dark exposure. The middle portion of the leaves was cut into three 2-cm sections. Top and bottom leaf sections were immediately snapped frozen in liquid nitrogen and used later for metabolite quantification. The middle leaf section was used in measurements of O₂-based R_dark, followed by determination of absolute soluble sugar and starch content (see below).

Photosynthesis and dark respiration rates

Leaves harvested from 6 individual plants per species were measured for rates of O₂-based R_dark at 30 °C using a fluorophore oxygen sensor (Astec Global, Maarssen, The Netherlands) as described in O’Leary et al. (2017) and Scafaro et al. (2017). Area and fresh mass of leaves were recorded, and then leaves were dried for at least 2 days at 60 °C to determine dry mass, soluble sugar, and starch contents.

Rates of CO₂-based R_dark and light-saturated photosynthesis (A_sat) in intact leaves were quantified using a LI-COR 6400-XT infrared gas analyzer (Li-Cor BioSciences, Lincoln, NE, USA) in a separate set of plants, at midday and midnight. At midday, leaves from 3 to 4 individual plants per species were used to quantify A_sat at a PPFD of 1,500 µmol quanta m⁻² s⁻¹, with the LI-COR chamber temperature and sample CO₂ concentration being set to 30 °C and 400 ppm, respectively. Following the A_sat measurements, the same leaves were wrapped in aluminum foil to dark-adjust for 30 min. CO₂-based R_dark was then measured using the same conditions as the A_sat measurements, but with the light turned off. At midnight, leaves from another set of 3 to 4 individual plants per species were measured for CO₂-based R_dark as described above. Leaf sections used in the LI-COR measurements were cut and dried for at least 2 days at 60 °C to determine dry mass. The apparent RQ was calculated using the CO₂-to-O₂ ratio of R_dark (average values or individual values; see text). Note that 2 separate techniques were used to measure CO₂-based and O₂-based R_dark and this could impact on RQ values, and as such, we refer to report values as “apparent” rather than “absolute” RQ values. The validity of our measurements of RQ is assessed in Supplementary Note 2.

Starch and soluble sugar analysis

Oven-dried leaf sections that were initially used in O₂-based R_dark measurements were ground to fine powder, and 5 to 10 mg of the powder was placed in a 2-mL microfuge tube with 0.5 mL of 80% (v/v) ethanol to extract soluble sugar and starch. The tissue was vigorously vortexed for 20 s and incubated in a Thermomixer orbital shaker (Eppendorf South Pacific Pty. Ltd.), set at 80 °C for 20 min with 1 × g shaking. The tissue was then centrifuged for 5 min at 16,260 × g, and the supernatant was collected. The extraction steps above were repeated on the pellet twice, and the supernatant was pooled. The pooled supernatant was used for determination of soluble sugars using a Fructose Assay Kit (catalog #FA20-1KT; Sigma-Aldrich Inc.) and invertase from baker's yeast (Saccharomyces cerevisiae; catalog #I4504; Sigma-Aldrich Inc.), while the remaining pellet was used for determination of starch using a Total Starch Assay Kit (catalog #K-TSTA-100A; Megazyme Inc.), following manufacturer's instructions and Rashid et al. (2020). A standard curve of soluble sugar was generated using a series of known concentrations of sucrose, glucose, and fructose stocks (Sigma-Aldrich Inc.). Absorbance was recorded using a microplate reader (Infinite M1000Pro; Tecan Group Ltd.) at 340 nm for sugars or 515 nm for starch.

GC-MS metabolite analysis

GC-MS was used to quantify metabolite profiles of leaves harvested at midday and midnight. Metabolites were extracted according to the procedure described in Howell et al. (2009) and Che-Othman et al. (2020) with some modifications. Frozen leaf tissue was ground to fine powder and approximately 25 mg of the powder was transferred into a 2-mL microfuge tube. Then 0.5 mL of cold extraction solvent mix with internal standards (1:2.5:1 (v/v/v) chloroform, methanol and water, 0.1% (v/v) L-Valine-¹³C₆ and D-Sorbitol-¹³C₆) was added. The mixture was vortexed (15 min, 4 °C) and centrifuged at 14,500 × g (15 min, 4 °C) for a total of 3 times, and the 3 supernatants were pooled. 0.4 mL HPLC-grade water was added to separate the phases, and the upper phase was collected and dried in a SpeedVac vacuum concentrator (Thermo Fisher Scientific Inc.) at 30 °C overnight. Chemical derivatization was performed using the MPS2 XL-Twister autosampler (Gerstel GmbH & Co. KG, Mülheim an der Ruhr, Germany) and samples were analyzed with an Agilent GC/MSD system comprising an Agilent GC 6890N (Agilent Technologies). 1 μ L of derivatized sample was injected (at 250 °C injector temperature) in split-less mode at a purge flow rate of 50 mL min⁻¹. Helium was used as the carrier gas at a flow rate of 1 mL min⁻¹. Compounds were eluted using the following temperature gradient: hold for 1 min at 70 °C then ramp at 7 °C min⁻¹ to 325 °C and finally hold for 3.5 min. The ion transfer line was heated to 280 °C and the ion source and quadrupole were heated at 150 and 230 °C, respectively. The resulting peaks were analyzed using MS-DIAL software v.4.48 (http://prime.psc.riken.jp/compms/msdial/main.html). The height of the quantifier ion (quantifying mass) for each peak was compared between samples after normalization. Metabolites were normalized against the averaged signals of 2 internal standards and leaf fresh mass, followed by weighing against the average measured signal across all samples for each compound (i.e. z-score normalized).

Statistical analysis

Gas-exchange and carbohydrate concentration data were analyzed using 2-way ANOVA and linear mixed-effect models. Metabolic responses to time of the day were examined using PCA, hierarchical clustering (i.e. heatmap) and multiple comparison tests. These tests were performed using R v.4.1.1 (R Core Team 2018). Comparisons were significant if P < 0.05. All data were checked with Bartlett's test for linearity, normality and heteroscedasticity. R packages used for PCA and hierarchical clustering were factoextra, FactoMineR, NbClust, cluster and pheatmap. z-Scored normalized metabolite levels were further log-transformed before performing the PCA. Relationships between metabolite and R_dark data were examined by OPLS as described by Cui et al. (2019), with metabolite concentrations as predicting variables (X) and R_dark or RQ as a predicted response variables (Y). OPLS tests were performed by SIMCA 17 (Umetrics, Umea, Sweden). The goodness of fit was examined using the correlation coefficient between observed and predicted Y (R²) and its cross-validated value (Q²), and the P-value associated with the difference between a random model (average $±$ random error) and the OPLS model (this P-value is referred to as P_CV-ANOVA). Results are also presented as volcano plots combining the output of the OPLS (loading on the x axis) and −log(P) obtained via the ANOVA (sample classes) or the regression (quantitative response variable). Best drivers of R_dark or RQ thus appear at the extremities of the volcano plot.

Accession numbers

No sequence data is generated in this article.

Acknowledgments

We thank Associate Prof. Oula Ghannoum and Dr Florence Danila for providing seeds. We are also grateful to staff of the ANU Research School of Biology Plant Services Team for maintaining the plants in the controlled environments.

Author contributions

Y.F., R.T.F., S.v.C., G.T., and O.K.A. planned and designed the study. Y.F. and N.L.T. conducted the experiments. All authors interpreted the data. Y.F., A.P.S., and O.K.A. wrote the first draft. G.T. improved the metabolic analyses and contributed to the second draft substantially. All authors contributed to the final version.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. Barplots of relative abundance (mean $±$ SE) of 47 metabolites in eight C₃ and C₄ grasses at midday and midnight.

Supplementary Figure S2. Biplots of principal component analysis (PCA) portraying relationships among dark-exposed metabolites in examined C₃ and C₄ species.

Supplementary Figure S3. Score plot of multivariate OPLS analysis of metabolites taking respiratory quotient as an objective response variable, colored by sampling timepoints (i.e. midday, midnight).

Supplementary Figure S4. Midnight respiratory quotient (RQ) versus concentration ratios of malate and aspartate in C₃ and C₄ dark-adapted leaves.

Supplementary Figure S5. Summary of sugar metabolism involving glucose 1-phosphate, myoinositol, and glycerate.

Supplementary Data Set 1. Data of gas-exchange, carbohydrates and metabolomes measured in eight C₃ and C₄ grasses at midday and midnight.

Supplementary Note 1. Malate catabolism in C₃ leaves.

Supplementary Note 2. Assessment of our technique to measure apparent RQ values.

Funding

This work was funded by grants from the Australian Research Council and was supported by the ARC Centre of Excellence in Plant Energy Biology (CE140100008), and the ARC Centre of Excellence for Translational Photosynthesis (CE1401000015). Y.F. was supported by the ANU International PhD Scholarship (737/2018) and HDR Fee Remission Merit Scholarship.

Data availability

The data that support the findings of this study are available within the Supplementary Data of this article.

References

Author notes